#spatial
2019-11-27
Stephanie Hicks (08:43:42): > @Stephanie Hicks has joined the channel
Stephanie Hicks (08:43:42): > set the channel description: Discuss data infrastructure and analysis methods for spatial data in Bioconductor
Stephanie Hicks (08:48:34): > set the channel topic: discuss data infrastructure and analysis methods for spatial data in Bioconductor
Kevin Rue-Albrecht (08:48:39): > @Kevin Rue-Albrecht has joined the channel
Alan O’C (08:49:27): > @Alan O’C has joined the channel
Charlotte Soneson (08:49:51): > @Charlotte Soneson has joined the channel
Kasper D. Hansen (08:52:08): > @Kasper D. Hansen has joined the channel
Casey Greene (08:55:38): > @Casey Greene has joined the channel
Martin Morgan (09:07:36): > @Martin Morgan has joined the channel
Izaskun Mallona (09:07:54): > @Izaskun Mallona has joined the channel
Frederick Tan (09:12:15): > @Frederick Tan has joined the channel
Sean Davis (09:29:33): > @Sean Davis has joined the channel
Stephanie Hicks (09:45:01): > welcome everyone!
Stephanie Hicks (09:46:20): > Ok. So let’s talk about how we should structure our working group
Stephanie Hicks (09:46:30): > e.g. should we start a google doc or github repo where we could start tracking packages?
Stephanie Hicks (09:49:11): > e.g. should we talk about data infrastructures that have been developed to store spatial data. one that I’m aware is the starfish schema (https://spacetx-starfish.readthedocs.io/en/latest/getting_started/index.html) from@Ambrose Carr@Justin Kigginsat the CZI. I think a nice goal for our working group could be to implement a version of this schema in R/Bioc to kick around.
Stephanie Hicks (09:53:53): > or maybe someone in the working group has already done that?
Kasper D. Hansen (09:57:21): > Frankly, I think we need some examples
Kasper D. Hansen (09:58:03): > I think trying to design a single class that fits everything is probably not a good idea. We should start by sketching out some examples and then we can later see if we can have a design which captures all needs
Kasper D. Hansen (09:59:11): > We also need to think carefully about the difference between say images and processed locations. Historically, for example, SummarizedExperiment is really for processed data with the idea that multiple different raw data representations could be transformed into the same processed data
Kasper D. Hansen (09:59:31): > I don’t have enough actual spatial experience to say what the bestprocessedrepresentation is
Kasper D. Hansen (09:59:57): > The more “raw” the data is, the less likely it is to have a common representation
Kasper D. Hansen (10:01:26): > The starfish representation looks to me to be pretty raw data. I mean, that’s also useful, but that’s different from processed data
Kasper D. Hansen (10:02:26): > Its not really clear to me that you have expression counts (ie a matrix) when you have data at the starfish level
Dario Righelli (10:03:14): > @Dario Righelli has joined the channel
Kevin Rue-Albrecht (10:06:24): > > The starfish representation looks to me to be pretty raw data. I mean, that’s also useful, but that’s different from processed data > Indeed, perhaps more than a representation of therawdata, we could discuss and identify the structure(s) that would make the most sense to use for downstream analyses,a la SummarizedExperiment
Kevin Rue-Albrecht (10:12:34): > > I don’t have enough actual spatial experience to say what the bestprocessedrepresentation is > Same thing here. But speaking of sketches, I wonder if a class derived fromSingleCellExperimentwith a “SpatialDims” slot - similar to reducedDims - could be a solution: > > @spatialDims > | | x | y | z | ... | > | cell | | | | | >
Koen Van den Berge (10:20:49): > @Koen Van den Berge has joined the channel
Laurent Gatto (10:28:07): > @Laurent Gatto has joined the channel
Charlotte Soneson (10:30:17): > Spaniel(https://bioconductor.org/packages/Spaniel/) seems to store the processed data (count matrix) from spatial transcriptomics in aSingleCellExperimentobject, with x/y coordinates in thecolData. The image is then read in separately.
Davide Risso (10:32:40): > @Davide Risso has joined the channel
Stephanie Hicks (10:33:45): > there is also Giottohttps://www.biorxiv.org/content/10.1101/701680v1
Davide Risso (10:34:46): > I second what@Kasper D. Hansensaid… we need examples and exemplar datasets to be able to understand what we need from an infrastructure point of view
Stephanie Hicks (10:34:46): > code:http://spatial.rc.fas.harvard.edu
Kevin Rue-Albrecht (10:36:30): > Good startSpaniel, though looking at the code > 1. I’d drop the “Depends: Seurat” (and drop support for Seurat altogether - users can convert their object) > 2. formally define a slot rather than storing as metadata (to enable validity checks) > 3. like SCE, make a package dedicated to the data structure, and leaving plotting functions to downstream packages
Davide Risso (10:36:42): > @Levi Waldronand I (and his student), as part of a somewhat broader project on single-cell multi-omics, have started to list some popular protocols and exemplar datasets with the idea of creating an ExperimentHub data package
Levi Waldron (10:36:46): > @Levi Waldron has joined the channel
Davide Risso (10:36:58): > Not sure how this overlaps with the effort of this group
Davide Risso (10:37:42): > @Stephanie Hicksshould we try to define a bit the scope of this group? Are we talking about spatial data per se or spatial + transcriptomic data only?
Stephanie Hicks (10:37:50): > to be clear, I alsothink getting our hands on data is the best way forward
Kasper D. Hansen (10:38:17): > My long point is that different technologies are likely to generate different types of output. For example, I can gurantee that the (x,y) representation of a cell is not relevant for all types of data. That’s not a problem, but it would be good to have some idea of how different techs are mapped into a structure
Kasper D. Hansen (10:40:12): > For example if you are interested in spatial proteomics and have sub-cellular resolutioon and you want to map proteins to different cellular compartments (which is a use case - perhaps very specialized) you can’t get away with assinging a cell to a (x,y) point
Davide Risso (10:41:27) (in thread): > Sorry, didn’t see the thread in#general, I guess we want to at least cover spatial transcriptomics and spatial proteomics
Kasper D. Hansen (10:43:37): > I’m curious about which technologies map to the (x,y, optional z) coordinate + expression matrix which Giotto uses. That seems to be a relatively easy case to start by handling. My example with sub-cellular resolution is likely to be much more complex
Kevin Rue-Albrecht (10:45:51) (in thread): > Here is an (x,y,z) example :https://www.biorxiv.org/content/10.1101/765842v1
Koen Van den Berge (10:48:07): > Yes I agree, the first big distinction seems to be expression data for cells from which you know the location vs having coordinates of gene expression as derived by e.g. fluorescence techniques (e.g. seqFISH+, ) where you have subcellular resolution (and may or may not know the boundaries of cells in that coordinate space, here it might not make sense to have a genes x cells matrix).
Koen Van den Berge (10:48:30): > Should we maybe start a Google Doc to classify technologies based on what we expect their data looks like?
Charlotte Soneson (10:51:18): > For an overview, perhaps this presentation from a student of Sten Linnarsson (from a course we organized last month) is useful:https://nbisweden.github.io/single-cell_sib_scilifelab/session-outlook/Spatial_Lars-Borm_NBIS2019.pdf
Davide Risso (10:54:55): > I don’t know much about spatial transcriptomics, but presumably there are preprocessing methods that infer cell boundaries and assign fluorescence to cells?
Davide Risso (10:55:55): > So I guess the distinction is whether someone needs / wants to have sub-cellular resolution or is fine with cellular resolution
Kasper D. Hansen (10:57:41): > There are almost guranteed to be interesting applications where you don’t care about cell boundaries but just look at spatial domains of expression. The bottom example on the giotto page falls into that category
Kasper D. Hansen (10:58:44): > getting cell boundaries could be pretty difficult for some approaches I think, and might be kind of irrelevant, depending on question
Davide Risso (10:59:34): > sure! But there are also examples in which all you care is how cells are organized within a complex tissue. What I’m saying is that there might not be a one-fits-all data structure but that we may want to think at different structures for different problems/applications
Kasper D. Hansen (11:00:11): > yes, and we are hashing out that
Kasper D. Hansen (11:00:31): > I might add that you might not need individual cells for organization
Kasper D. Hansen (11:00:43): > That is what I was trying to say
Davide Risso (11:04:22): > I think we agree, I didn’t mean to sound like we don’t…:slightly_smiling_face:Perhaps because I mostly work with brain data these days, the first example in the presentation that@Charlotte Sonesonposted is what I had in mind
Davide Risso (11:05:41): > (e.g. slides 19 to 29)
darlanminussi (11:06:39): > @darlanminussi has joined the channel
Justin Kiggins (11:10:08): > @Justin Kiggins has joined the channel
Justin Kiggins (11:16:07): > for examples, I highly recommend@Ambrose Carr’s prototype athttps://starspace.readthedocs.io/notable: > * a schema for a spatially-localized gene expression matrix that supports with transcriptomics and proteomics and spatial sequencing methods > * examples of 9 public datasets converted into the schema > https://starspace.readthedocs.io/en/latest/conversion_examples/index.html
Ambrose Carr (11:16:12): > @Ambrose Carr has joined the channel
kipper fletez-brant (11:30:07): > @kipper fletez-brant has joined the channel
Justin Kiggins (11:30:24) (in thread): > :point_up:this. for starfish, we’d like to make sure that it spits out processed data that downstream tools can use WITHOUT the analyst-scientist needing to go back to the images
Justin Kiggins (11:31:40) (in thread): > this is consistent with how we’ve been coercing spatial data to load into cellxgene via h5ad… by treating the spatial dimensions like tSNE or UMA
Kevin Rue-Albrecht (11:33:53) (in thread): > Cool. > Well, all I’m saying here is that a dedicated slot would clarify the spatial nature of those coordinates, as opposed to reduced dimensions that don’t have the same spatial meaning
Justin Kiggins (11:34:31) (in thread): > starfish takes data from the image level to the processed gene expression matrix level. example of starfish’s outputs:https://spacetx-starfish.readthedocs.io/en/latest/help_and_reference/working_with_starfish_outputs/index.html
Justin Kiggins (11:34:46) (in thread): > yeah, totally
Martin Morgan (11:37:17) (in thread): > an interface tohttps://spacetx-starfish.readthedocs.io/en/latest/help_and_reference/spacetx-format/output_formats/index.htmlseems like some pretty low-hanging fruit and accessible with existing tools… RStarfish or StarfishExperiment depending on how close it is to features x samples
Justin Kiggins (11:44:07) (in thread): > this is really really good:smile:
Kasper D. Hansen (11:50:27) (in thread): > We should do something like that as part of our CZI grant for sure, since Starfish is CZI-developed
Justin Kiggins (11:58:37): > Also: Hi folks:smile:Thanks for the invite,@Stephanie Hicks! > > We’ve been thinking a lot about this a lot over the past few months & I just want to second & expand on some of points that have been made in this channel today already…
Justin Kiggins (12:01:30): > Standard representations of the processed data (spatially-annotated gene expression matrices) are the biggest need and the easiest place to find commonality across the many assays.
Justin Kiggins (12:04:12): > Subcellular features are important, at the very least for inspection and QC for the gene expression matrices. Our experience with starfish indicates that it should be possible to standardize the decoded spot calls across the image-based transcriptomics methods
Justin Kiggins (12:05:01): > annotating x/y/z coordinates of a tissue sample onto a gene expression matrix is pretty straightforward. we also have a sense that there are other spatial features that biologists analyzing these data want, like summary descriptions of cell morphology (size, shape, etc), quality score, etc. Then, as you aggregate across multiple tissue samples, x/y/z becomes meaningless & you need secondary spatial labels, like region labels or common coordinate framework. the details here will depend on tissue and species norms
Simina Boca (12:22:48): > @Simina Boca has joined the channel
Simina Boca (12:27:35): > Hi and thanks for starting the channel@Stephanie Hicks!
Ambrose Carr (12:27:44) (in thread): > +1, excellent summary of image-based and sequencing-based transcriptomics methods. A few links to some missing approaches that I think are interesting: > > insta-seq:https://www.biorxiv.org/content/10.1101/722819v1bar-seq:https://www.biorxiv.org/content/10.1101/294637v2There are also a few proteomics methods worth examining. > > 4i:https://www.ncbi.nlm.nih.gov/pubmed/30072512cycif:https://www.cycif.org/codex:https://www.akoyabio.com/application/files/7015/6625/6771/CODEX_Brochure_Aug_2019_WEB.pdf
Simina Boca (12:28:05): > I’ve started looking into some spatial things a bit, but we’ve just used this packagehttps://akoyabio.github.io/phenoptr/ - Attachment (akoyabio.github.io): inForm Helper Functions > Helper functions for working with inForm data.
Kevin Rue-Albrecht (12:42:57) (in thread): > I’m thinking > - SpatialCellExperiment > - SpatialFeatureExperiment > (As in, 2 subclasses for two purposes, not to choose between them)
Ambrose Carr (12:45:33): > @Stephanie Hickshere’s the use case summary from my literature review that prompted the prototype that@Justin Kigginslinked.https://docs.google.com/spreadsheets/d/1tLTJrGul6p_4dpiT1_uU46HFCj-EE2mABXwSxqfCryQ/edit?usp=sharingNote that there is some conflation between “I didn’t look” and “I couldn’t load the data”, so in cases where there’s a conspicuous lack of annotation for a study, chances are it’s the former.
Ambrose Carr (12:48:31) (in thread): > It should provide a sense of the current variability of the field, and substantiate the need for flexible abstractions for how coordinate spaces are represented.
Nitesh Turaga (13:09:56): > @Nitesh Turaga has joined the channel
Daniel Van Twisk (13:09:57): > @Daniel Van Twisk has joined the channel
Hervé Pagès (13:12:03): > @Hervé Pagès has joined the channel
Lori Shepherd (13:12:07): > @Lori Shepherd has joined the channel
Kayla Interdonato (13:12:37): > @Kayla Interdonato has joined the channel
JiefeiWang (13:18:06): > @JiefeiWang has joined the channel
Aedin Culhane (13:36:42): > @Aedin Culhane has joined the channel
Aedin Culhane (13:42:49): > Hi Ruben Dries gave a talk on spatial transcriptomics with Giotto at the Boston Bioconductor meetup. It includes a review of different spatial approaches. His slides are athttps://www.dropbox.com/s/4e0a9qrk8zhm900/191106_meetup_spatial_transcriptomics_small.pdf?dl=0
Lukas Weber (13:43:50): > @Lukas Weber has joined the channel
hcorrada (14:16:37): > @hcorrada has joined the channel
Leonardo Collado Torres (15:28:46): > @Leonardo Collado Torres has joined the channel
Andrew Jaffe (16:34:13): > @Andrew Jaffe has joined the channel
Andrew Jaffe (16:43:53): > we’ve been generating data from various flavors of spatial transcriptomics technologies for a bit now, and i’m happy to share our thoughts
Peter Hickey (18:47:27): > @Peter Hickey has joined the channel
2019-11-28
Helena L. Crowell (05:57:26): > @Helena L. Crowell has joined the channel
Martin Morgan (15:53:34) (in thread): > The ExperimentMatrix can be exported to loom, and then presumably imported to Bioconductor LoomExperiment so there is that much already…
Shila Ghazanfar (19:43:53): > @Shila Ghazanfar has joined the channel
2019-11-29
Vince Carey (09:13:32): > @Vince Carey has joined the channel
Kevin Rue-Albrecht (12:43:38) (in thread): > As mentioned a few times, real toy data sets will provide the use cases needed to inspire and test implementations. > > With that in mind, in the meantime, I’ve put together a straightforward prototypehttps://github.com/kevinrue/SpatialCellExperiment- borrowing heavily fromSingleCellExperimentcode - adding aspatialDimslot that behaves similarly to a singlereduceDimelement (cells can have multiple reduced dimension results, but a single set of real coordinates). > > I’ve put a couple of thoughts for other representations in the vignette, e.g.SpatialPolygonExperiment,SpatialTileExperiment#FunFriday
2019-12-01
Mikhail Dozmorov (10:16:04): > @Mikhail Dozmorov has joined the channel
Norbert Tavares (14:28:01): > @Norbert Tavares has joined the channel
2019-12-02
Stephanie Hicks (09:27:32) (in thread): > Thanks@Kevin Rue-Albrecht! This is great. This was exactly what I planned to today (if no one else had tried this over the US holiday break). I will give it a go with using one of the datasets@Ambrose Carrlinked us to last week
Kevin Rue-Albrecht (09:31:52) (in thread): > Happy to hear feedback about missing or future feature, and comparisons to existing solutions (e.g., Python). > In this first implementation, my main focus was to have the built-in subsetting of spatial coordinates along assay and metadata using the[syntax. That came straight out of SCE, using the internal metadata slot.
Martin Morgan (10:44:43) (in thread): > FWIW I hope to have some code to access / use zarr archives (via Python) from R later today…
Stephanie Hicks (11:36:13) (in thread): > oh that’s great, thanks@Martin Morgan!
Andrew Jaffe (14:42:43): > i can weigh in with a few comments…
Andrew Jaffe (14:42:52): > first i dont think its a good idea to assume any of the spatial transcriptomics data is at single cell resolution, particularly not with the commercially-available visium only have 55um spot resolution.
Andrew Jaffe (14:43:22): > other approaches like slideseq and emerging approaches could also be subcellular resolution, like several micron
Andrew Jaffe (14:43:31): > which would also not result in anything at the single cell level
Andrew Jaffe (14:44:41): > it would probably be easier to extend existing single cell objects to just have a spatial barcode as the sample instead of cell, and then try to have addition phenotype/sample information
Andrew Jaffe (14:45:08): > second, you probably want some kind of container for the histology image itself, particularly for visium
Andrew Jaffe (14:45:45): > then the sample/barcode information can have coordinates to overlay on the histology
Andrew Jaffe (14:46:50): > its also possible to count how many cells are in each spot/at each barcode although its not super straightforward, which you could append to the sample data…this would likely be computed outside of R since R isnt great with image analysis, although maybe that could change over time
Andrew Jaffe (14:47:51): > third, i also think it also might be hard for the microscopy-based approaches and sequencing-based approaches to have common data formats since the data are pretty different
Charlotte Soneson (17:55:10): > 10x Genomics have released a Visium spatial gene expression data set:https://support.10xgenomics.com/spatial-gene-expression/datasets/1.0.0/V1_Adult_Mouse_Brain
Aedin Culhane (17:56:59): > We prepared a mRNA (seqFISH) and proteomics (MIBI) spatial data challenges for the BIRS meeting. R rda fils are availablehttps://github.com/BIRSBiointegration/Hackathon
Stephanie Hicks (22:59:16): > Thanks@Andrew Jaffe— could you elaborate more on point #3 in terms of what you see as the major differences between microscopy-based and sequencing-based data formats?
Andrew Jaffe (23:00:32): > well, the data processing is totally different
Andrew Jaffe (23:00:49): > that type of data is more likely to actually be single cell resolution
Andrew Jaffe (23:01:13): > since there are likely processing steps that result in a cell by gene matrix with spatial information
Andrew Jaffe (23:01:39): > that i assume would be done way before reading the data into some bioc class
Stephanie Hicks (23:01:49): > ah ok i see what you mean
Martin Morgan (23:52:44) (in thread): > Please givehttps://github.com/Bioconductor/ZarrExperimenta try; the README provides installation instructions (more complicated than average) and walks through a use case… Would be good to get the@Aaron Luninsight into installing via basilisk
2019-12-03
Ruizhu HUANG (03:26:57): > @Ruizhu HUANG has joined the channel
Charlotte Soneson (04:07:10): > Without having actually looked at any real spatial data yet, following the comments above I was wondering if theTreeSummarizedExperimentcontainer (https://bioconductor.org/packages/TreeSummarizedExperiment/) could be useful to extend/adapt for this purpose (it already extendsSingleCellExperiment). Perhaps it’s not really a tree structure, but one could maybe imagine seeing the relation between pixels and cells as a grouping relation (several pixels per cell, or several cells per pixel), which could potentially be captured in the row/column link tables of the TSE.
Stephanie Hicks (06:52:56): > nice idea@Charlotte Soneson!
Stephanie Hicks (06:59:41): > so I started a repo to explore various spatial datasets (https://github.com/stephaniehicks/spatialexperiment-analysis). Using@Charlotte Soneson’s suggestion, I took the Visium v1.0.0 data (mouse brain) (.h5file for feature x barcode,tissue_positions_list.csvfor spatial coordinates, and just loaded the image files separately temporarily), and created aSpatialCellExperiment(https://github.com/kevinrue/SpatialCellExperiment) object via the GitHub R package that@Kevin Rue-Albrechtcreated:https://github.com/stephaniehicks/spatialexperiment-analysis/blob/master/data-analyses/visium-mouse-brain-strain-C57BL6.Rmd. Haven’t had a chance to do anything except create the object. Will work more on this today. Thoughts are welcome.
Stephanie Hicks (07:00:30): > Some of my initial thoughts: > * Might be better to useSpatialBarcodeExperimentfor this type of data (when it’s not necessarily at the cell level and you just have barcodes) or even justSpatialExperimentto be inclusive to truly cell or just barcoded data instead ofSpatialCellExperiment? > * Following up on@Andrew Jaffe’s suggestion, we definitely want to have a container for the histology image itself (or maybe we read in the RBG coordinates via thelibrary(jpeg)package? I tried to do this in my code. I can imagine various approaches to this and would love other people’s insights. Details on images returned from Visium herehttps://support.10xgenomics.com/spatial-gene-expression/software/pipelines/latest/output/images > * For Visium, you get two sets of spatial coordinates in thetissue_positions_list.csvfile (1) where the spots are located on the array and (2) where the spots are located as a column, row pixel coordinate in the image. We should discuss where to store both sets of coordinates. Would be good to discuss standardized naming conventions (or just keep as is?) Would be good to store metadata on the type of protocol used (e.g. Visium) so that the bioc object knows what to expect.
Stephanie Hicks (07:08:47): > the code uses thelibrary(here), so assuming one clones the repo with the.Rproj, it should work fine on anyone’s computer (aka b/c it uses relative paths). thoughts are welcome.
Kevin Rue-Albrecht (07:12:20): > I think we all agree that there is a need for multiple representations (i.e. classes) of spatial data, specialized for the various steps of the processing workflow (like processing microarrays from probes to genes). > > WithSpatialCellExperiment, I intentionally started at the very end of such a workflow, from an end-user perspective, assuming that: > - the histology image is already segmented into cells (i.e., polygons) > - cells are already summarized into a point coordinate (for better or worse, e.g. centroid) > - probes are already summarized per cell (i.e., segmented polygon) > > I agree that classes representing “raw-er” forms of the data - like the SpatialBarcodeExperiment that@Stephanie Hicksjust mentioned - are much needed for the upstream processing steps.
Simina Boca (08:52:05): > @Kevin Rue-AlbrechtIn SpatialCellExperiment, do the data have to be whole exome? I have an example that I think matches what you’re saying but only has 4 targeted proteins.
Kevin Rue-Albrecht (08:57:19): > Let me know if I misunderstand your question. > Just like the parentSummarizedExperiment, there is no restriction on the type of features that theSpatialCellExperimentcan contain. Those can be genes, transcripts, proteins, or even genetic variants. > There can be a few features, a single one, or the whole exome/transcriptome/proteome/…
Lukas Weber (09:07:50): > I agree it makes sense to generalize it e.g.SpatialExperimentorSpatialBarcodeExperiment, so in the default case we would have spatial barcodes. For the simpler case where the resolution is equal to cells (which will probably be a lot of cases in practice), maybe it makes sense to have a simple column in thecolDataindicating whether a barcode is equal to a cell, e.g. a logical vectoris_cellor a factor with some levels e.g.cell,barcode. (On the other hand, maybe this would be implicit in the rest of thecolData). > > Something like this could maybe also work for the sub-cellular high-resolution case – e.g. simply a column incolDataidentifying which cell each barcode belongs to. For lower-resolution cases (more than one cell per spatial barcode) I’m not sure. For more complicated cases I agree something like theTreeSummarizedExperimentcould be useful, although I’m not sure I fully understand this option yet. > > Would we ever expect sub-cellular resolution cases where one barcode matches to more than one cell, e.g. overlapping cells? Or would this usually be removed during the initial cell segmentation steps?
Simina Boca (09:17:31) (in thread): > Cool! A collaborator has multiplex IHC data for 4 proteins. We’ve used the phenotpR package with it, but it would be cool to try SpatialCellExperiment. That being said, the student working on just left, so I’ll see if I can get someone else.
Kevin Rue-Albrecht (09:20:18) (in thread): > Thanks for your interest! > That said I need to point out thatSpatialCellExperimentliterally just implements a candidate container (i.e. class) for the data. The package doesn’t offer any processing function. > I’ll add that I would highly encourage such “processing” functions to be implemented in downstream packages, to separate the core container from downstream analyses/workflows.
Simina Boca (09:21:31) (in thread): > I see! It inherits things from SummarizedExperiment though, right? Which I also need to look into:joy:
Kevin Rue-Albrecht (09:23:17) (in thread): > Yup. That’s the main advantage: it get the automatic synchronised subsetting of assays, colData, rowData, reducedDim, and spatialDim for free (i.e., when you subset the object, all those slots are subsetted in the output object)
Simina Boca (09:24:29) (in thread): > If other people here are interested in working on it, I can ask my collaborator too. It’s unpublished data so have to see if she’s OK with sharing.
Martin Morgan (09:28:02): > Want to mentionhttps://bioconductor.org/packages/EBImagefor working with images@Mike Smith@Wolfgang Huber - Attachment (Bioconductor): EBImage > EBImage provides general purpose functionality for image processing and analysis. In the context of (high-throughput) microscopy-based cellular assays, EBImage offers tools to segment cells and extract quantitative cellular descriptors. This allows the automation of such tasks using the R programming language and facilitates the use of other tools in the R environment for signal processing, statistical modeling, machine learning and visualization with image data.
Kevin Rue-Albrecht (09:29:03): > @Lukas WeberLast point (subcellular mapping to multiple cells): I’d be curious to see the difference between the two typical approaches: 1) remove ambiguous assignments (like: mapped reads, doublets), 2) attempt to deconvolute based on other unambiguous assignments (like: pseudoaligned reads) > > First point, do I see a bit of contradiction? > > For the simpler case where the resolution is equal to cells > > a simple column in the colData indicating whether a barcode is equal to a cell > If the resolution is equal to cells, there is no need to indicate if a barcode is equal to cell, right? > > I think a separate class for barcode-level dataSpatialBarcodeExperimentwould probably make the most sense, until the data is actually segmented into individual cells. > For that, I’m intrigued by@Charlotte Soneson’s idea about the possibility of extending TreeSummarizedExperiment, which could iteratively group barcodes, possibly until the cell level.
Simina Boca (09:29:14): > I’ve used that EBImage before and like it!
Simina Boca (09:29:44): > It does get a bit slow if you have hundreds of files though
Martin Morgan (09:31:48) (in thread): > Seems like often the number of images in these experiments is usually small?
Simina Boca (09:40:34) (in thread): > Yeah, I was using it for images from a kaggle contest where you have to train/test models on hundreds of files. It’s super user friendly though and I really like it!
Stephanie Hicks (09:47:05) (in thread): > there are like 6-8 images in the visium data that I was analyzing above
Lukas Weber (10:13:57) (in thread): > Thanks, yes this makes sense. So then you would suggest having two classes: the simple case where we have cell level, and a more general class based on barcodes, which possibly builds onTreeSummarizedExperiment.
Kevin Rue-Albrecht (10:14:53) (in thread): > exactly
Lukas Weber (10:15:33) (in thread): > And thanks for setting up theSpatialCellExperimentprototype! Looks great so far
Kevin Rue-Albrecht (10:18:20) (in thread): > FWIW, in the vignette ofSpatialCellExp, I’ve also suggested aSpatialPolygonExpwhich could be half-way between the histology image and the over-simplistic point representation of cells that is SpatialCellExp. > i.e., each cell could be represented by asf(CRAN) polygon, a representation than lighter image but more accurate than a centroid point
Lukas Weber (10:20:13) (in thread): > Ok, interesting, I’ll have a look. I’m not familiar withsffrom CRAN
Kevin Rue-Albrecht (10:22:55) (in thread): > I’ mostly aware of it from the useR 2018 sticker wall, where it’s used to draw a country map
Kevin Rue-Albrecht (10:22:56) (in thread): > https://www.mitchelloharawild.com/blog/user-2018-feature-wall/ - Attachment (mitchelloharawild.com): useR! 2018 feature wall > Creating the hexmap feature wall for useR! 2018
Hervé Pagès (12:44:40): > That makes a lot of sense@Charlotte Soneson. Sounds almost inevitable that at some point people will want to perform some kind of hierarchical clustering of the cells or the pixels of an image. Some minor adjustments to the TreeSummarizedExperiment container would be needed though since right now, IIRC, it is geared towards phylogenetic trees. Alternatively, if the goal is to support more generic trees, maybe we should consider moving the tree capabilities down the stack e.g. to SingleCellExperiment or even to RangedSummarizedExperiment or SummarizedExperiment.
Jenny Drnevich (13:53:51): > @Jenny Drnevich has joined the channel
Vince Carey (15:25:03) (in thread): > Just FYI@Aaron Lun@Martin MorganI will take a crack at the basilisk approach in a fork.
Aaron Lun (15:25:15): > @Aaron Lun has joined the channel
Vince Carey (15:39:12) (in thread): > It is proving difficult to install numcodecs on my mac.
Aaron Lun (15:57:32): > Wait, what? Why do we want trees to live deeper in the stack? Seems like you could just wrap any tree in a column-parallel class and stick in in thecolDatasomewhere.
Andrew Jaffe (16:09:46): > EBImage is very good for reading in and small manipulations of images, but it doesnt have the same extent of operations that exist in more imaging centric programs like matlab
Andrew Jaffe (16:10:08): > like we do a lot of image processing in matlab and the read in processed data to R
Aaron Lun (16:17:54) (in thread): > Is this a basilisk problem?
Vince Carey (16:25:02) (in thread): > no, it is a mac problem. the basilisk setup is simple …
Aaron Lun (16:40:36): > Without having looked at the data, it seems to me that for the visium stuff, each “sample” (i.e., spot) is associated with an x- and y- axis coordinate. You can stick these in a SE as usual. To preserve spatial information… if only we had some way of representing pairs of coordinates… > > Oh wait, we do. It’s the********GenomicInteractions**** ****class. Comes with 1D and 2D overlap methods, bounding boxes, and other bits and pieces. If you rephrase the coordinates from nucleotides to some integral unit of distance, you’re good to go.
Hervé Pagès (19:17:19) (in thread): > For the same reason that we have a slot & dedicated getter and setter for therowRanges? It’s good that we have the flexibility to put all kind of stuff inrowDataand/orcolDatabut there are also benefits in using formal slots + accessors for things that we want to “standardize”. So for example downstream code can trust that something likecolTree(se)will return a tree-like object (or NULL) so doesn’t need to do any guess work.
Vince Carey (20:43:59) (in thread): > https://github.com/vjcitn/ZarrExperimentis a fork and associated PR illustrating basilisk provisioning of python components.
Aaron Lun (20:44:55) (in thread): > was the numcodecs installation triky?
Vince Carey (21:00:37) (in thread): > There was no problem in the warm embrace of bioconductor/bioconductor_full:devel, but i have been completely stymied on mac and from the looks of the issues on the githubhttps://github.com/zarr-developers/numcodecs/issues/210i am not alone
Vince Carey (21:03:11) (in thread): > In the docker container vjcitn/zarrexp:v3 you can install ZarrExperiment, but need to install rmarkdown additionally to build the vignette.
2019-12-04
Aaron Lun (00:09:53): > On another note, if we want to have vectorized spatial operations, we should probably make something likeRgeoslib. sfdepends on system installation of GEOS, which is hardly the definition of “user-friendly”.
Charlotte Soneson (13:26:08): > Space Ranger (10x Genomics software for analyzing spatial gene expression data) is now available:https://support.10xgenomics.com/spatial-gene-expression/software/overview/welcome
2019-12-05
Marcel Ramos Pérez (16:56:58): > @Marcel Ramos Pérez has joined the channel
Aedin Culhane (17:21:56): > Playing with visualizing MIBI images.. What are others using? - File (Plain Text): rspatial.R snippet
Aedin Culhane (17:31:39): > @Aaron Lungeojson is ok to work with but the shape (.shp) files is better for layered graphics. There are nice extensions to ggplot for shp files, tm_ functions work well tmap::tm_shape, etc. I played around with GIS and geojson with our local town maps. Happy to send you maps of the last local elections!!! The files athttps://mgimond.github.io/Spatial/reading-and-writing-spatial-data-in-r.htmland library(sf) were easy to work with - Attachment (mgimond.github.io): Reading and writing spatial data in R | Intro to GIS and Spatial Analysis > This is a compilation of lecture notes that accompany my Intro to GIS and Spatial Analysis course.
Aedin Culhane (17:38:14): > its easy to convert geojson to shp
Aaron Lun (17:38:51): > I don’t care so much about the file format right now. I basically just want vectorized polygon operations for in-memory manipulation of summarized data (cell boundaries, tissue features, etc.), mimicking our GenomicRanges infrastructure. It seems likesfdoes this, but that package depends on three system libraries that do not seem like they would be regular installations.
Aedin Culhane (17:39:54): > see tmap funcitons
Aedin Culhane (17:40:07): > I think sf and tmap work tightly together
Aaron Lun (17:40:30): > Yes, I know, the question is what we need to do above and/or below it to make it easier to install/use.
Aedin Culhane (17:44:39): - File (Plain Text): Untitled
Aedin Culhane (17:46:25): > I am happy to push the data/code to git.. It was just playing around with GIS data, because I thought it would be useful for spatial stuff down the road
Aedin Culhane (17:47:26): > ( I made a shiny app of ithttps://firefly.shinyapps.io/NewtonVotes/)
Aedin Culhane (17:48:10): > Within the shp file you can define multiple polygons, sub polygon etc, so its makes it easier to do regional and sub regional analysis
Aedin Culhane (17:48:39): > However I didn’t create these shp files. I pulled them from the city.
Aedin Culhane (17:51:23): > my next to-do was getting MIBI tiff files to work with sf and tmap. But I didn’t get there. I expected other might have a better solution that I
Aedin Culhane (17:52:32): > What formats are the CZI/HCA pushing, I saw some talk of geojson but I wasn’t sure which way it was going?
Aedin Culhane (18:48:58): > Hi Aaron, will this help
Aedin Culhane (18:49:27): - File (Plain Text): Untitled
Aedin Culhane (18:50:44): > Visualization of MIBI image contours using tmap - File (PNG): Screen Shot 2019-12-05 at 6.49.49 PM.png
Aedin Culhane (18:51:35): > Visualization of same tiff file as a RasterBrick - File (PNG): Screen Shot 2019-12-05 at 6.50.51 PM.png
Aedin Culhane (18:51:52): > Data is available as SpatialDataFrame, and I can access the xy coordinates and values
Aedin Culhane (18:52:39): - File (R): Untitled
Aaron Lun (18:53:55): > Let me clarify my question. I have no doubt thatsf,tmapand associated packages will do the job. My question is whether it is okay to expect users to installGDAL (>= 2.0.1), GEOS (>= 3.4.0), PROJ (>= 4.8.0)themselves. This seems like a high barrier to entry, and “just use containers” is not a real answer for most of our users.
Aedin Culhane (18:55:40): > Aaron, sorry to misunderstand you. I can’t answer that question. Why have these such a burden, is it because they include lat/long maps?
Aedin Culhane (18:58:47): > Is there an existing slim package with just the functionality we need. I was excited to be able to get the xy coordinates and matrix out. For analysis of these data, we need to be able to re-create a psudo image, and/or project results onto a background image.
Aaron Lun (19:04:05): > Well, that’s the thing. The GEOS C++ seems to have exactly what we would want, and it seems pretty battle-hardened: > > Geometries: Point, LineString, Polygon, MultiPoint, MultiLineString, MultiPolygon, GeometryCollection > > Predicates: Intersects, Touches, Disjoint, Crosses, Within, Contains, Overlaps, Equals, Covers > > Operations: Union, Distance, Intersection, Symmetric Difference, Convex Hull, Envelope, Buffer, Simplify, Polygon Assembly, Valid, Area, Length, > So I can understand whysfdepends on it. Any replacement would probably struggle to beat the range and efficiency of available operations.
Aedin Culhane (19:06:24): > if a large body of existing standards exist, its easier to build and leverage them,
Aaron Lun (19:07:04): > Which leads us to our current problem.
Aedin Culhane (19:07:22): > Is there an alternative to GEOS C++ . or slim version
Aedin Culhane (19:22:34): > Looking at the HCA CZI slack channel, many seem to prefer ZARR format. I see martin and yourself have already got a Zarr experiment package. What’s the difference in these formats?
Hervé Pagès (20:02:55): > FWIW I don’t think depending onsfwould be a problem. It’s already installed on all the build machines (which indicates that some Bioconductor package already depends directly or indirectly on it) except on celaya2 but that’s only because CRAN doesn’t provide Mac binaries for R 4.0 yet. Although this is a temporary situation for which the workaround is to install the binaries for R 3.6 (I just did this so now sf is also on celaya2).
Kasper D. Hansen (20:04:20): > Im having some issues with sf on our cluster, because of dependencies. And because I want a maintainable solution, and I therefore don’t want to handcompile stuff
Kasper D. Hansen (20:04:56): > I also havn’t really dived into it, but Im just echoing@Aaron Lunthat it is not trivial.
Hervé Pagès (20:38:37): > Yeah on my laptop where I still have Ubuntu 16.04 LTS, I run into this: > > > install("sf") > Bioconductor version 3.11 (BiocManager 1.30.10), R Under development (unstable) > (2019-10-30 r77336) > Installing package(s) 'sf' > trying URL '[https://cran.rstudio.com/src/contrib/sf_0.8-0.tar.gz](https://cran.rstudio.com/src/contrib/sf_0.8-0.tar.gz)' > Content type 'application/x-gzip' length 8607770 bytes (8.2 MB) > ================================================== > downloaded 8.2 MB > > * installing **source** package 'sf' ... > **** package 'sf' successfully unpacked and MD5 sums checked > **** using staged installation > configure: CC: gcc > configure: CXX: g++ -std=gnu++11 > checking for gdal-config... /usr/bin/gdal-config > checking gdal-config usability... yes > configure: GDAL: 1.11.3 > checking GDAL version >= 2.0.1... no > configure: error: sf is not compatible with GDAL versions below 2.0.1 > ERROR: configuration failed for package 'sf' > * removing '/home/hpages/R/R-4.0.r77336/library/sf' > > The downloaded source packages are in > '/tmp/RtmptgwTK9/downloaded_packages' > Updating HTML index of packages in '.Library' > Making 'packages.html' ... done > Warning message: > In install.packages(...) : > installation of package 'sf' had non-zero exit status > > However the build machines have Ubuntu 18.04 LTS which comes with a version of GDAL that satisfies fs requirements. So yes, people running old Linux distros won’t be able to install sf out-of-the-box.
2019-12-06
Aaron Lun (11:19:06): > The most interventionalist solution would be to make a Rgdallib, Rprojlib and Rgeoslib to package up those requirements, see if CRAN would be willing to build them, and then see if all relevant packages (sfetc.) would be willing to link to the resulting libraries.
Aaron Lun (11:20:02): > This would be the most robust solution but relies on a lot of people sitting down and singing kumbaya next to a campfire, so I don’t think it’s going to happen.
Aaron Lun (11:23:36): > A slightly less intrusive version of this idea would be to tell users to set appropriate environment variables to the Rgeoslib install location, and hope thatsf’sconfigurefiles can pick them up. Requires a bit of work from the users but… well, at least it’s now inconvenient rather than impossible.
Aaron Lun (11:25:17): > At the other extreme, we can roll our own bindings to these libraries. This may or may not be something desirable, depending on how usablesf’s interface is.
Hervé Pagès (11:46:19): > FWIW.deband.rpmpackages for GDAL (>= 2.0.1), GEOS (>= 3.4.0) and Proj.4 (>= 4.8.0), are available for old Ubuntu and Fedora distros. Seehttps://github.com/r-spatial/sf#linuxI just installed them on my laptop (Ubuntu 16.04 LTS “xenial”) to replace the “official” ones (which are too old forsf) then was able toinstall.packages("sf").
2019-12-08
Mark Robinson (14:41:21): > @Mark Robinson has joined the channel
2019-12-10
Chris Vanderaa (09:32:10): > @Chris Vanderaa has joined the channel
Olly Crook (09:34:29): > @Olly Crook has joined the channel
Davide Risso (11:14:03): > Posting here too - Attachment (satijalab.org): Satija Lab > Lab Webpage —
Davide Risso (11:15:00): - Attachment: Attachment > 10x has shared some great dataset as well for their visium protocol https://www.10xgenomics.com/solutions/spatial-gene-expression/
Ludwig Geistlinger (11:15:24): > @Ludwig Geistlinger has joined the channel
Avi Srivastava (11:16:04): > @Avi Srivastava has joined the channel
Kevin Blighe (11:20:42): > @Kevin Blighe has joined the channel
Aaron Lun (11:35:28): > I saw aSpatialExperimentclass somewhere. Do we . hvae a repo?
Avi Srivastava (11:39:07): > https://github.com/kevinrue/SpatialCellExperimentthis ? I am not completely sure I am still going through previous chats.
Aaron Lun (11:40:10): > @Kevin Rue-AlbrechtThat should probably not have “Cell” in the title.
Avi Srivastava (11:48:33): > Interesting, I was following the discussion and I am wondering should we try and understand the Seurat object first? > like it seems through the tutorial they are processing both the image and the coordinates, then why re invent the wheel, just use that, no ?
Stephanie Hicks (11:48:49): > @Aaron LunI agree (https://community-bioc.slack.com/archives/CR1BLV821/p1575374430123600), but based on my understanding of@Kevin Rue-Albrecht’s response he was arguing to have “raw-er” and more “processed” forms of the data and would need different containers for such. - Attachment: Attachment > Some of my initial thoughts: > • Might be better to use SpatialBarcodeExperiment for this type of data (when it’s not necessarily at the cell level and you just have barcodes) or even just SpatialExperiment to be inclusive to truly cell or just barcoded data instead of SpatialCellExperiment ? > • Following up on @Andrew Jaffe’s suggestion, we definitely want to have a container for the histology image itself (or maybe we read in the RBG coordinates via the library(jpeg) package? I tried to do this in my code. I can imagine various approaches to this and would love other people’s insights. Details on images returned from Visium here https://support.10xgenomics.com/spatial-gene-expression/software/pipelines/latest/output/images > • For Visium, you get two sets of spatial coordinates in the tissue_positions_list.csv file (1) where the spots are located on the array and (2) where the spots are located as a column, row pixel coordinate in the image. We should discuss where to store both sets of coordinates. Would be good to discuss standardized naming conventions (or just keep as is?) Would be good to store metadata on the type of protocol used (e.g. Visium) so that the bioc object knows what to expect.
Aaron Lun (11:49:12) (in thread): > Because I don’t want to depend on the Seurat object.
Aaron Lun (11:49:50): > Life would have been a lot easier if they had derived from an SE subclass, but they didn’t, and now none of their stuff works with our stuff. So there you have it.
Aaron Lun (11:50:58): > I’m counting from v3 here, v2 was a mess that is better left untalked about.
Avi Srivastava (11:51:11): > well no offense , it is sounding more like a competition:stuck_out_tongue_winking_eye:It goes both ways, I was not involve in either project so I shouldn’t comment much here but I was curious.
Aaron Lun (11:51:53): > Sometimes I dream about what life would be like if they used the SCE class. I could just outsource all my plotting functions to them, and that would make life so much easier for me.
Aaron Lun (11:52:48): > But, much like world peace and solving global hunger, it’s just a dream.
Avi Srivastava (11:53:34): > :slightly_smiling_face:I am guessing you already tried talking to them ? but I think there might me more to that then just changing the object.
Aaron Lun (11:53:43): > Yes, there is. They should be on BIoC.
Avi Srivastava (11:54:07): > gotcha
Kevin Rue-Albrecht (11:54:22) (in thread): > I can remove the “Cell” in this package. Depends whether we want to store the various classes in this one package, or implement each class in a different package, a-la-SCE
Avi Srivastava (11:54:54): > Anyways, I am excited to see so many people at one place interested in spatial sequencing data.
Kevin Rue-Albrecht (11:54:59) (in thread): > +1
Aaron Lun (11:55:28) (in thread): > We should some kind of SpatialVector (maybe already provided by the SpatialDF) and we can just subclass the SCE and SE to support that being parallel along the columns.
Avi Srivastava (11:55:48): > Just wanted to add into the mix, I am able to use alevin on 10x spatial data as well as on the slide-seq data
Avi Srivastava (11:56:10): > Please feel free to ping me if anyone is interested from the quantification side .
Camille BONAMY (11:57:08): > @Camille BONAMY has joined the channel
Aaron Lun (11:57:43) (in thread): > I should probably elaborate on this. The major practical difficulty of depending on a non-BioC package is that their changes take place outside of the BioC release cycle, so it is hard to guarantee stability in release.
Alan O’C (11:58:25) (in thread): > Speaking of reinventing the wheel in a discussion of Seurat certainly has an element of irony to it
Aaron Lun (11:58:39) (in thread): > Case in point is the recent update to Rcpp; IIRC, this caused some strife as all downstream packages needed to be actively version bumped (at the very least) to trigger recompilation against the new ABI.
Avi Srivastava (11:59:23) (in thread): > My apologies, I didn’t mean to offend anyone, I was just curious and asked.
Alan O’C (11:59:25) (in thread): > Maybe I’m missing something, but why does one need to distinguish betweenSpatialBarcode, andSpatialCell? What difference between the objects would there be? Is there more information beyond “spot/cell” co-ordinates and “spot/cell” metadata in either case?
Aaron Lun (11:59:50) (in thread): > Now, we tolerate that because I trust the Rcpp authors to do solid engineering and to preserve back compatibility. I also don’t mind depending on CRAN packages with limited liability where I can easily patch over breaking changes.
Aaron Lun (12:01:46) (in thread): > For Seurat, I lost that trust with v2->v3 where the data structure completely changed and some functions just disappeared. (The data structure change wouldn’t have been so bad by itself, but they were recommending direct slot access before, so a lot of analysis code was broken.)
Avi Srivastava (12:02:03) (in thread): > I think a spot can have multiple cells, not sure if metadata already has it but in future tools might deconvolve it.
Aaron Lun (12:02:59) (in thread): > It was also kind of annoying because their v3 object is pretty close to an SCE. So why didn’t they just use that instead and save everyone some time? When@Davide Rissosuggested that to Rahul in 2017, the negative response was motivated by the desire to not break code. Ok, fair enough at the time, but the code is broken now anyway.
Alan O’C (12:03:00) (in thread): > Yes, but why do you need to distinguish between a barcode/spot and a cell? Is there a circumstance for which the data structure for each of these use cases needs to be different?
Aaron Lun (12:03:20) (in thread): > Anyway. That’s my rant completed.
Avi Srivastava (12:03:50) (in thread): > Thanks for giving the field view Aaron, appreciate that.
Aaron Lun (12:04:51) (in thread): > Just a future that could have been.:chuuni:
Kevin Rue-Albrecht (12:07:33) (in thread): > Happy to receive PRs on SpatialCE if that fits in there. I’m not doing much with the package at the moment
Alan O’C (12:12:17) (in thread): > In the case when you’re aggregating barcodes to a cell, there’s maybe a meaningful distinction. But you could also aggregate to (eg) tissue level. In that case storing a vector of counts per tissue and a corresponding bounding box would be more flexible
Alan O’C (12:15:45) (in thread): > I’m certainly not offended! I’ve been annoyed by changes in SCE as much as changes in Seurat:slightly_smiling_face:
Aaron Lun (12:16:29) (in thread): > Hey, at least user-facing code never broke. Or at least had 6 months of deprecation.
Alan O’C (12:17:53) (in thread): > I never claimed my annoyance was justified, it’s just nice to have someone to blame
Almut (12:31:02): > @Almut has joined the channel
Hervé Pagès (14:18:02) (in thread): > > It was also kind of annoying because their v3 object is pretty close to an SCE > Only good news here is that it should now make it easier to coerce back and forth between seurat v3 object and SCE. Then all your plotting functions could just doplot(as(SCE, "seurat v3")). A dream come true!
Hervé Pagès (14:20:48) (in thread): > Until they change the object again with seurat v4:grin:
2019-12-11
Aaron Lun (01:02:56): > Anyway, moving on from that.@Hervé Pagèsit seems like installation ofsfis (just) tolerable enough to allow us to use it as some sort of thing-that-is-parallel-to-the-columns of a SE subclass.
Elisabetta Mereu (06:09:14): > @Elisabetta Mereu has joined the channel
Hervé Pagès (11:45:08): > @Aaron LunSeems to work: > > library(sf) > nc <- st_read(system.file("gpkg", "nc.gpkg", package="sf"), quiet=TRUE) > library(SummarizedExperiment) > se <- SummarizedExperiment(matrix(runif(800), ncol=100)) > colData(se) <- DataFrame(nc=I(nc)) > plot(colData(se)$nc["SID74"]) > > The fact that you need to wrap the sf object inI()is annoying and error prone though…
Aaron Lun (11:45:51): > Guess we could just throw a lite S4 wrapper to get it into theVectorhierarchy.
Hervé Pagès (11:47:50): > I guess so. Or ask the sf folks to switch to S4. Just kidding…
Somesh (15:24:02): > @Somesh has joined the channel
Aedin Culhane (15:37:25) (in thread): > spatial data will have >2 dimensions, starmap have a xyz coordinates.
Kevin Rue-Albrecht (15:42:49) (in thread): > I haven’t enforced a number of coordinates inSpatialCellExperiment. The slot expects a matrix with any number of columns (dimensions). Rows of the spatial dimensions correspond tocolnames(se)
2019-12-13
Kevin Rue-Albrecht (09:57:20): > #FunFriday > For abbreviations purposes, I might have stumbled onto a solution to disambiguate the existing SCE acronym from the spatial one: Spatial Single-Cell Experiments (SSCE). > Credits tohttps://github.com/hubmapconsortium/vitessce
Michael Steinbaugh (11:15:56): > @Michael Steinbaugh has joined the channel
Lukas Weber (12:41:00): > New method for combining spatial and single-cell data, posted on bioRxiv today:https://www.biorxiv.org/content/10.1101/2019.12.13.874495v1
Lukas Weber (12:41:22): > Available as a Python package calledstereoscope
Lukas Weber (12:42:03): > I think it is from the groups in Sweden who were originally involved with developing the earlier versions of the Visium technology
Aaron Lun (12:43:18): > Smells like good ol’ deconvolution to me.
2019-12-16
Stephanie Hicks (21:57:03): > https://aws.amazon.com/blogs/industries/building-a-scalable-image-processing-pipeline-for-image-based-transcriptomics/ - Attachment (Amazon Web Services): Building a scalable image processing pipeline for image-based transcriptomics | Amazon Web Services > The below post was written by guest author Shannon Axelrod, Senior Software Engineer at the Chan Zuckerberg Initiative. At the Chan Zuckerberg Initiative (CZI), the science team’s goal is to support the science and technology that will make it possible to cure, prevent, or manage all diseases by the end of this century. In particular, the organization […]
Aaron Lun (23:08:35): > If this basilisk thing ever gets off the ground, one of my first pet projects would have been to wrap starfish.
Ellis Patrick (23:47:23): > @Ellis Patrick has joined the channel
2019-12-17
Jean Yang (02:42:12): > @Jean Yang has joined the channel
Stephanie Hicks (09:45:35) (in thread): > That would be amazing if you (or someone else) would be willing do that:slightly_smiling_face:
Martin Morgan (12:25:24) (in thread): > Not sure what the ultimate goal is here, but it is fairly easy to use starfish from R using reticulate. I cloned the repository > > git clone git@github.com:spacetx/starfish.git > cd starfish > > then created and activated a python virtual environment > > virtualenv -p python3 ~/.virtualenvs/starfish > source ~/.virtualenvs/starfish/bin/activate > > then installed the python requirements > > pip install -r REQUIREMENTS.txt > > and started R with the reticulate package and imported the starfish module > > $ bioc-devel > > library(reticulate) > > starfish = import("starfish") > > Taking a look at one of the python notebooks from starfish (https://github.com/spacetx/starfish/blob/master/notebooks/BaristaSeq.ipynb), I started down the analysis path… > > > experiment_json = paste0("[https://d2nhj9g34unfro.cloudfront.net/browse/formatted/20190319/baristaseq](https://d2nhj9g34unfro.cloudfront.net/browse/formatted/20190319/baristaseq)", "/experiment.json") > > exp = starfish$Experiment$from_json(experiment_json) > > exp > <starfish.Experiment (FOVs=25)> > { > fov_000: <starfish.FieldOfView> > Primary Image: <slicedimage.TileSet (z: 17, r: 7, c: 4, x: 913, y: 1193)> > Auxiliary Images: > dots: <slicedimage.TileSet (z: 17, r: 1, c: 1, x: 913, y: 1193)> > fov_001: <starfish.FieldOfView> > Primary Image: <slicedimage.TileSet (z: 17, r: 7, c: 4, x: 913, y: 1193)> > Auxiliary Images: > dots: <slicedimage.TileSet (z: 17, r: 1, c: 1, x: 913, y: 1193)> > fov_002: <starfish.FieldOfView> > Primary Image: <slicedimage.TileSet (z: 17, r: 7, c: 4, x: 913, y: 1193)> > Auxiliary Images: > dots: <slicedimage.TileSet (z: 17, r: 1, c: 1, x: 913, y: 1193)> > fov_003: <starfish.FieldOfView> > Primary Image: <slicedimage.TileSet (z: 17, r: 7, c: 4, x: 913, y: 1193)> > Auxiliary Images: > dots: <slicedimage.TileSet (z: 17, r: 1, c: 1, x: 913, y: 1193)> > ..., > } > > > > a hint is to useBiocManager::install("rstudio/reticulate")to get tab completion onstarfish$Exp<tab>. > > I wonder what a more complete ‘wrapper’ would look like (and I guess in the end what the benefit would be, since it seems like one wants to interact with starfish basically as python, rather than python from R…). I guess I’m lacking imagination at the moment…
Aaron Lun (12:49:15) (in thread): > Mostly because it would have been fun to try. I don’t really have any motivation beyond that.
Martin Morgan (13:24:41) (in thread): > it could still be fun! especially perhaps identifying common analysis steps that can be bundled into R functions rather than notebook scripts…
Stephanie Hicks (14:38:38) (in thread): > thanks@Martin Morgan!
2019-12-18
Nick Canete (19:31:22): > @Nick Canete has joined the channel
2020-01-13
Friederike Dündar (11:29:54): > @Friederike Dündar has joined the channel
2020-02-05
Charlotte Soneson (03:21:46): > scanpy now handles Visium data (storing the counts, images and spatial coordinates in the AnnData object):https://nbviewer.jupyter.org/github/theislab/scanpy-tutorials/blob/master/analysis-visualization-spatial.ipynb - Attachment (nbviewer.jupyter.org): Notebook on nbviewer > Check out this Jupyter notebook!
Aaron Lun (03:28:47): > Sheesh. We’re always late to the party.
Aaron Lun (03:31:54): > Well, whatever. As soon as basilisk gets off the ground, tu casa es mi casa for all Python stuff as far as I’m concerned.
Friederike Dündar (14:55:48): > I’m using visium data right now in SCE, just added the xy coordinates as a reducedDim
Friederike Dündar (14:56:48): > works fine for my purposes, I don’t need to drag the tissue image along all the time anyway
Friederike Dündar (15:40:44): > what would be the goal for BioC? have a standardized object for the tissue slide?
Aaron Lun (15:41:26): > I don’t care so much about the slide itself (though we should certainly consider it). I mostly care about the features derived from the slide, represented as polygons.
Aaron Lun (15:41:46): > The aim would be to be able to do something likefindOverlaps(cells, tumor)and get the same stuff that we get for ranges.
Friederike Dündar (15:42:04): > why polygons?
Aaron Lun (15:42:07): > Or even stuff likefindOverlaps(cells, tumor, maxgap=1)e.g. for all cells within 1 micron of tumors.
Aaron Lun (15:42:17): > Or points, depending on the resolution.
Aaron Lun (15:42:38): > I just mention polygons as a default but we can also take the centroid.
Friederike Dündar (15:42:59): > I don’t think any of the packages can handle that, but in principle it comes down to annotating the spots first anyway
Friederike Dündar (15:43:09): > and then they can be used just like any old cluster/cell annotation, no?
Aaron Lun (15:43:28): > Hold on, hold on. There are several different concepts here.
Friederike Dündar (15:43:40): > yes, I’m just talking about Visium for now
Aaron Lun (15:43:59): > Right. There’s the spot for visium, which is a multi-cell thing. Then there’s other technologies, where we get segmented cells. Then we could be thinking about individual transcript spots.
Friederike Dündar (15:44:06): > yes
Aaron Lun (15:44:16): > Visium is actually the easiest of the lot because the spots are in a grid with the same radius.
Friederike Dündar (15:44:23): > scanpy doesn’t handle that either
Friederike Dündar (15:44:29): > yes, Visium is easiest
Friederike Dündar (15:44:46): > and my argument would be that BioC can absolutely handle Visium data
Aaron Lun (15:45:06): > Right now, yes. Having said that, it may still be valuable to approximate the diameter of the spot with a polygon, because then you could ask questions like “give me all spots that have at least 50% overlap with my feature of interest”.
Aaron Lun (15:45:30): > In any case, it’s not too much effort to support both polygons and points in this new architecture. I have already spec’d out how these overlaps would work.
Aaron Lun (15:45:43): > I JUST NEED (WO)MANPOWER.
Aaron Lun (15:46:14): > One of the bad things about working in industry is that we can’t just spin up a grad student to work on a problem.
Friederike Dündar (15:46:30): > you could get an intern
Aaron Lun (15:46:45): > Harder than you would think.
Aaron Lun (15:46:51): > And they don’t stick around.
Aaron Lun (15:47:18): > @Shila Ghazanfarshould join this channel if she isn’t in here already.
Aaron Lun (15:47:22): > Oh, I guess she is.
Shila Ghazanfar (15:58:21): > I’ve been lurking, you’ve blown my cover:sleuth_or_spy:
Shila Ghazanfar (16:06:43): > i think framing in terms of polygons is certainly worthwhile, despite a lot of data out there being at a lower resolution than single-cell, there is a decent amount of higher resolution techniques (FISH based, in situ seq?) and not just in transcriptomics either (think IMC). I think a good first step is to get from a segmentation mask from e.g. ilastik and get a set of polygons & associated features per cell, and in the same coordinate system as the transcriptomic data
Friederike Dündar (16:10:54): > if your goal is to play catch up with Theis and Satija, then demonstrating how to handle Visium data would be a first start, since most people will have access to it given that it’s commercially available
Aaron Lun (16:11:02): > y’know@Friederike Dündarif you’ve got a running analysis can you post it somewhere?
Friederike Dündar (16:11:25): > I will
Friederike Dündar (16:11:39): > do you have a preferred spot?
Aaron Lun (16:11:49): > Do you?
Friederike Dündar (16:12:02): > nope
Aaron Lun (16:12:08): > Maybe GH pages then
Aaron Lun (16:13:51): > Anyway, time-wise there’s two separate aspects; getting data structures and getting an analysis up and running.
Aaron Lun (16:14:13): > I don’t really care much about the latter - people can play fast and loose for now - but I do want to get a data structure into the next BioC release.
Aaron Lun (16:15:46): > Or to be precise: there are and will be a million ways to do the analysis to get scientific insights, and that’s fine. But they should all use the same data structure so that you can seamlessly switch between them, and that requires some serious thought and careful engineering.
Kasper D. Hansen (16:16:28): > doing something with (x,y) seems sensible, but depending on the assay that may not capture the info
Aaron Lun (16:17:10): > And you know I love cheap grad student labor, but the engineering is probably something that would require more oversight.
Aaron Lun (16:38:09): > I mean, we have a few months until the next release anyway, so we won’t be getting into production before then.
Friederike Dündar (21:22:14): > what do you want to provide the data structure for? One container to bind them all? Or a separate entity for the image?
Aaron Lun (22:34:11): > I would like a structure to represent spatial features.sp actually already provides them, so the aim is to just wrap those and make them work with BioC conventions, e.g.,findOverlapsand so on. Mayberesizeandreduce, etc.
2020-02-06
Shila Ghazanfar (06:25:37): > Similar tofindOverlapsI’d like something likefindNeighboursfor any n’th degree, with cells sharing edges of the polygonalisation
Friederike Dündar (09:53:01): > :+1:
Friederike Dündar (10:29:31): > the spatstat package may have some functionalities that could be useful, too
Friederike Dündar (10:29:45): > spatstat.org/resources/spatstatQuickref.pdf
Friederike Dündar (10:30:23): > > For windows, several polygon operations are de- scribed to produce new windows, and how to subset the point pattern to a redefined boundary. Operations on pixel images are also described and how to produce heat maps with them.
cakapourani (16:38:55): > @cakapourani has joined the channel
2020-02-08
Sara Ballouz (05:53:01): > @Sara Ballouz has joined the channel
2020-02-10
Domenick Braccia (14:49:39): > @Domenick Braccia has joined the channel
2020-02-11
Theresa Alexander (15:35:04): > @Theresa Alexander has joined the channel
2020-02-12
Davide Risso (06:03:10): > Has anyone already started a github repo or similar to start to hash out these ideas?
Davide Risso (06:04:28): > We’re also very interested in getting this off the ground (I believe@Stephanie Hicksand@Kasper D. Hansenare interested too) and now that@Dario Righelliworks with me we might have the man power that@Aaron Lunwas looking for…
Davide Risso (06:05:00): > Unless@Aaron Lunalready implemented everything (which is always a possibility!):smile:
Davide Risso (06:05:34): > should we start by limiting the scope to just visium data for a first package?
Davide Risso (06:08:22): > something like (i) basic data structure (inheriting from SCE), (ii) simple methods (accessors, findOverlaps and the like), (iii) simple visualization methods (image colored by cluster / UMI count / etc.)?
Kevin Rue-Albrecht (06:18:48): > I don’t have the time to work on it for the foreseeable future, but speaking of inheriting from SCE, it’s any useful, I’m happy to hand overhttps://github.com/kevinrue/SpatialCellExperiment
Casey Greene (06:18:57): > Manubot manuscript?:grin:
Kevin Rue-Albrecht (06:22:52) (in thread): > Disclaimer: It’s only a concept prototype developed over a weekend. I never got a chance to apply it to real data. I believe@Stephanie Hickstook it for a test run at some point
Kevin Rue-Albrecht (06:25:11) (in thread): > There it was:https://community-bioc.slack.com/archives/CR1BLV821/p1575374381122700 - Attachment: Attachment > so I started a repo to explore various spatial datasets (https://github.com/stephaniehicks/spatialexperiment-analysis). Using @Charlotte Soneson’s suggestion, I took the Visium v1.0.0 data (mouse brain) (.h5 file for feature x barcode, tissue_positions_list.csv for spatial coordinates, and just loaded the image files separately temporarily), and created a SpatialCellExperiment (https://github.com/kevinrue/SpatialCellExperiment) object via the GitHub R package that @Kevin Rue-Albrecht created: https://github.com/stephaniehicks/spatialexperiment-analysis/blob/master/data-analyses/visium-mouse-brain-strain-C57BL6.Rmd. Haven’t had a chance to do anything except create the object. Will work more on this today. Thoughts are welcome.
Shila Ghazanfar (06:40:11): > I’ve worked a bit on transforming a pixel-by-pixel cell segmentation (in h5 format output e.g. from ilastik) into either 1. polygon vertices of the convex hull per cell, 2. polygon vertices based on the boundaries of the cell. There is a speed/accuracy trade-off here, convex hull much faster but gives poor results when the cell’s aren’t actually convex, and my boundary method implementation is a bit slower:grimacing:This is useful for visualisation, and works towards the data structure for doing polygon-related operations.. > > I should mention this is more relevant for the imaging based spatial gene expression datasets where individual mRNAs are spatially resolved (seqFISH/MERFISH etc..), but we should keep multiple data modes in mind, and nothing’s stopping one from using this polygonalisation at a lower resolution (think tissue classification based on the H&E) > > I wonder if anyone has previous experience with manipulating polygon/spatial data formats? e.g. using thesporsfpackages? Would be really helpful to pick your brain if so, thanks!!:slightly_smiling_face:
Davide Risso (08:20:51): > Thanks@Shila Ghazanfar! That’s awesome! Any public repo that we can contribute to? Or should we start on some orthogonal aspect and then merge?
Thanh Le Viet (09:42:32): > @Thanh Le Viet has joined the channel
2020-02-13
Aaron Lun (14:10:49): > I don’t have much to add right now, but if you’re thinking of a name for a package that is for single-cell FISH data analysis, you might considerscuttlefish.
Martin Morgan (19:25:35): > https://github.com/Bioconductor/Contributions/issues/1370is worth constructive comment
Ellis Patrick (20:52:33) (in thread): > There is way too much to comment on. But it does the job for what I need it to do.
2020-02-14
Hervé Pagès (03:13:50): > @Aaron Lunwhy not just cellfish?
Andrew Skelton (05:10:17): > @Andrew Skelton has joined the channel
Ellis Patrick (12:51:20) (in thread): > The name of my class is very misleading as it sounds like it extends SummarizedExperiment when it really is just a nested DataFrame. Any suggestions for a different name? I’ve been saying SegmentedCellExperiment for so long that my brains locked in. scDataFrame? spatialCells? imageDF
Alan O’C (12:57:48) (in thread): > Is there a practical reason for the transpose? I’d imagine you get a lot for free by storing data as in *Experiment classes
Ellis Patrick (17:10:36) (in thread): > I typically work with 10-40 proteins and 100s thousands or millions of cells. The microarray convention feels a bit icky. I was very tempted just to extend a flowSet but the rectangular structure of a nested DataFrame has just made it easier for me to conceptualise my data.
Alan O’C (17:16:55) (in thread): > I mean “icky” isn’t a practical reason, and I’d tend towards the convenience of having a whole bunch of existing methods over a better feel, but I’m a lazy programmer who’s already spent far too much time re-inventing Bioconductor-style functionality in a previous job. Having said that, I have no horse in this race, and you seem to have considered the options before going with it so you’re probably well justified:slightly_smiling_face:Was just curious
2020-02-15
Aaron Lun (14:40:23) (in thread): > I remember@Charlotte Sonesononce showed me a cytof dataset where the data was transposed - rows were cells, columns were markers - and it was stored in a SE! Nasty stuff; I spent a while trying to figure out what on earth was happening in the code that was being run on this object. The short of it is, if you putExperimentin the name of your data structure, the assumption is that the rows are features and the columns are samples.
Ellis Patrick (16:19:42) (in thread): > Definitely changing the name:slightly_smiling_face:. Also thinking about just rolling it into the analysis package that is using it too.
2020-02-16
Kasper D. Hansen (09:40:32): > Having looked a bit at this, it is pretty clear to me that we’re going to have 2 different classes, one for visium data where - because of the assay - we have a fixed (x,y) position for each measurement, and one for the image-based spatial expression that@Ellis Patrickand@Shila Ghazanfaris talking about. Those two classes are going to be quite different at first, I think.
Shila Ghazanfar (10:33:31): > Agreed, in my current analyses I have two separate objects reflecting the different resolutions, one for cell-level and another for spot-level. Keeping them consistent with each other worries me, so it may be worth designing a function that converts from the spot-level data into a SCE (maybe including cell polygon vertices), or some joint object class? SpatialDotExperiment and/or SpatialCellExperiment ?
2020-02-21
Dario Righelli (11:47:49): > Hi guys, I started working on a draft of a class for 10x Visium Spatial Transcriptomics:https://github.com/drighelli/VisiumExperiment
Dario Righelli (11:48:15): > I created also another repository with data and rmd file for testing:https://github.com/drighelli/VisiumAnalysis
Dario Righelli (11:48:51): > If any of you has any time for looking at it, or comments or whatever, please let me know.
Dario Righelli (11:51:52): > Keep in mind that it’s a first draft, so many things need to be improved and added. > Improvements apart, next step is the mapping of the barcodes clusters on the image spots. > If you have any package to suggest for this, I’d be very happy to try to use it. > At the moment I’m stucked on ggmap and lattice, with absolutely no knowledge of them.
Leonardo Collado Torres (16:53:52): > Hi! > > We are working on a new pkg that we’ll submit to BioC, potentially next week. The package will have: > * Ways to download the 12 Visium data that Andrew mentioned previously. > * The data is stored in a SingleCellExperiment object with the histology images stored insidemetadata(sce) > * Functions for plotting data from the data. > * The code for launching ashinyapp that lets you annotate spots to a given tissue/spot label of your choosing. > * Some other functions related to our particular dataset. > The idea is that it could be useful to you if: > * You simply want some more Visium data. > * Want to re-shape your data into what ours is structured as, then re-use the visualization functions and/or the shiny app itself. > * Want to explore our data in more detail. > I’ll let you know once we submit it to BioC so that everyone can access it. The shiny app is not meant to compete withiSEEand I guess thatiSEE(or other apps) could be updated to handle this type of data. It was designed with feedback from our biologists such that they could annotate the spots. Compared to Loupe from 10x Genomics, it’s a bit more bulky but it does let you explore multiple samples. > > Best, > Leo
Aaron Lun (17:43:32): > I missed the “not” in “not meant to compete” and my first thought was “WE WILL CRUSH YOU”.
Aaron Lun (17:45:39): > Anyway, it is now very easy to customize iSEE’s panels for arbitrary UI elements; you might consider making a PR into iSEEu with a VisiumPlot panel type.
Aaron Lun (17:45:49): > https://github.com/iSEE/iSEEU
Aaron Lun (19:17:17): > More generally, FTTB I think it’s good to have multiple objects. We’ll gather all the ideas into a single object at some point just like we did for the SCE.
Leonardo Collado Torres (23:44:37) (in thread): > hehehe, nah
Leonardo Collado Torres (23:45:53) (in thread): > I like the PR idea for further down the road.
Leonardo Collado Torres (23:46:16) (in thread): > Yup, I think of our pkg as a prototype where some pieces/ideas can be molded into the final BioC *Experiment pkg for this type of data as well as the explorer app.
Aaron Lun (23:53:12) (in thread): > Well, every man and his dog are making visualization apps these days, so the market is super crowded.
2020-02-24
Hervé Pagès (16:48:05) (in thread): > Note that you can have the convenience/simplicity of DataFrame AND the microarray layout at the same time by using (or extending) a TransposedDataFrame. See?TransposedDataFramein S4Vectors.
2020-02-25
Andrew Erickson (11:32:35): > @Andrew Erickson has joined the channel
2020-02-26
Leonardo Collado Torres (23:23:14): > Hi! Here’s what we promised last weekhttps://github.com/Bioconductor/Contributions/issues/1389. This is basically the pre-preprint as we are hoping (lots of:female-technologist::male-technologist::reverse_conga_parrot:) to submit to bioRxiv soon. Anyway, thanks for taking a look inside!
Leonardo Collado Torres (23:27:29): > I think some of you simply might be interested because this is harder than the example mouse data: the mouse brain is small enough that you can fit most of it on a Visium slide. Since brain regions are quite different, lots of clustering methods pick the differences and find the brain regions. But that’s harder in this scenario. Anyway, lots of work to do!
2020-02-27
USLACKBOT (10:03:16): > This message was deleted.
Leonardo Collado Torres (10:21:58) (in thread): > ?
Davide Risso (10:45:00) (in thread): > Access Denied: Project isb-cgc-01-0006: User does not have bigquery.jobs.list permission in project isb-cgc-01-0006.
Davide Risso (10:45:06) (in thread): > this is what I get…
Sean Davis (11:56:14) (in thread): > My bad. Wrong channel.
2020-02-28
Leonardo Collado Torres (19:00:28): > Hi again! I just wanted to let you know thathttps://github.com/LieberInstitute/HumanPilotis now public. That’s where we have our analysis code that includes my initial explorations withscranandzinbwave(though we then shifted towards pseudo-bulking and usinglimmafor some of our analyses). The README includes links to the raw data from 10x.
Leonardo Collado Torres (19:02:35): > we also deployed thespatialLIBDshiny app athttp://spatial.libd.org/spatialLIBD/and a few mirrors as shown athttp://spatial.libd.org/promp_error
Leonardo Collado Torres (19:12:53): > Also, if anyone wants to collaborate about taking parts of spatialLIBD and maybe building a software/infrastructure package, let me know.
2020-02-29
Leonardo Collado Torres (10:04:04) (in thread): > Thanks Kevin for noticing the typo haha
Leonardo Collado Torres (15:01:06): > https://support.bioconductor.org/p/128690/
2020-03-01
Kasper D. Hansen (20:41:14): > Well, the images should go out of$metadatain my opinion, but I appreciate the need for a quick hack
Kasper D. Hansen (20:41:25): > I would also make the images optional
Kasper D. Hansen (20:41:42): > (which they may be right now)
2020-03-19
Leonardo Collado Torres (14:30:25): > https://twitter.com/fellgernon/status/1240705824487403522?s=20is happening now - Attachment (twitter): Attachment > We are going live in 10 minutes! :scream::raised_hands::skin-tone-5: > > Join @10xGenomics, @kr_maynard and myself to learn more about our recent work, it’s free ^^ > > Useful links: > > :computer:http://research.libd.org/spatialLIBD/ > :scroll:https://www.biorxiv.org/content/10.1101/2020.02.28.969931v1 > > #SpatialTranscriptomics #Visium #DLPFC #ASD #SCZD #scRNAseq #Bioconductor https://twitter.com/TheScientistLLC/status/1238104331125231619 - Attachment (twitter): Attachment > [FREE Webinar] “Transcriptome-Scale Spatial Gene Expression in the Human Dorsolateral Prefrontal Cortex” Join us Thursday, March 19th at 2:30 PM ET. Register here: http://bit.ly/3aq1MTJ https://pbs.twimg.com/media/ES6hVAtWsAAFQX6.jpg
Leonardo Collado Torres (14:30:47): > I’ve never presented to 600 ppl before:sweat_smile:
Marcel Ramos Pérez (14:33:57): > The link sends me to a different event
Leonardo Collado Torres (14:35:42): > https://webinars.the-scientist.com/prefrontal-cortex-visium-spatial-10x?utm_source=Social_Media&utm_campaign=10x_Promotions%EF%BB%BF
Leonardo Collado Torres (14:36:30): > thx for letting me know!
2020-04-12
gcyuan (14:52:33): > @gcyuan has joined the channel
gcyuan (15:05:29): > Thanks@Vince Careyfor the invitation.
2020-04-14
Stephanie Hicks (16:07:01): > Saw this from@Avi Srivastavavia twitter and thought it might be relevant for others here —https://combine-lab.github.io/alevin-tutorial/2020/alevin-spatial/ - Attachment (combine-lab.github.io): Spatial Alevin > Spatial Single-Cell Quantification with alevin
2020-04-16
Al J Abadi (02:05:25): > @Al J Abadi has joined the channel
2020-04-20
Nils Eling (02:46:58): > @Nils Eling has joined the channel
2020-04-27
Aaron Lun (18:18:26): > @Leonardo Collado TorresI was going to test out spatialLIBD but it seems like it never built in BioC-devel.
Leonardo Collado Torres (18:20:57): > well, it was just recently accepted (I was late responding to messages by the reviewers). Anyway, I’m sure that it’ll be there in a day or 2. In the meantime, you can install it from GitHub. And you can check the docs athttp://research.libd.org/spatialLIBD/ - Attachment (research.libd.org): LIBD Visium spatial transcriptomics human pilot data inspector > Inspect interactively the spatial transcriptomics 10x Genomics Visium data from Maynard, Collado-Torres et al, 2020 analyzed by Lieber Institute for Brain Development researchers and collaborators.
Aaron Lun (18:22:01): > Oh, hold on, it doesn’t revdep SCE.
Aaron Lun (18:22:15): > I was wondering why I couldn’t find it when searching the SCE revdep list.
Leonardo Collado Torres (18:23:06): > https://github.com/LieberInstitute/spatialLIBD/blob/master/DESCRIPTION#L66
Aaron Lun (18:23:20): > Damn, your Depends are all the way at the bottom.
Aaron Lun (18:25:20): > Okay. Will look at your data structures in a second.
Leonardo Collado Torres (18:26:10): > http://research.libd.org/spatialLIBD/articles/spatialLIBD.html#re-shaping-your-data-to-our-structuremight be useful then - Attachment (research.libd.org): Introduction to spatialLIBD > spatialLIBD
Aaron Lun (18:27:23): > > ## Information about the image coordinates stored in imagerow and imagecol > colData(sce)[, c("imagerow", "imagecol")] > #> DataFrame with 47681 rows and 2 columns > #> imagerow imagecol > #> <numeric> <numeric> > #> AAACAACGAATAGTTC-1 113.141 147.435 > #> AAACAAGTATCTCCCA-1 383.438 413.051 > #> AAACAATCTACTAGCA-1 129.523 231.008 > #> AAACACCAATAACTGC-1 431.188 155.806 > #> AAACAGCTTTCAGAAG-1 344.869 125.068 > #> ... ... ... > #> TTGTTGTGTGTCAAGA-1 287.039 357.606 > #> TTGTTTCACATCCAGG-1 431.773 248.065 > #> TTGTTTCATTAGTCTA-1 442.259 210.801 > #> TTGTTTCCATACAACT-1 361.341 202.115 > #> TTGTTTGTGTAAATTC-1 157.021 277.993 > > Is all I wanted.
2020-05-04
Nitin Sharma (06:26:58): > @Nitin Sharma has joined the channel
2020-05-05
Aaron Lun (16:52:35): > @Hervé Pagèsyou may be interested inhttps://github.com/kevinrue/spatula
Aaron Lun (16:52:53): > Probably just jump to the vignette.
Aaron Lun (21:19:09): > @Leonardo Collado Torresare those layers manually annotated or derived from the expression data?
Aaron Lun (21:19:18): > Shocked if the latter, that would be ridiculously clean.
Aaron Lun (21:19:36): > Talking about the figure in Secition 6.1.
Aaron Lun (21:21:28): > Ah, I see it in 4.1.
Aaron Lun (21:30:17): > Two points come to mind: > * This dataset and your analysis could well serve as the basis for a book on using Bioconductor for spatial tx analyses. Separate from OSCA, that one can’t tolerate any more content. > * Did you use/do you want an unbiased feature selection tool? I’m thinking of slapping together something based on Moran’s I.
Andrew Jaffe (22:41:54): > they were manually annotated
Andrew Jaffe (22:42:05): > we’re happy to contribute to such a book
Andrew Jaffe (22:42:19): > can you expand that last point on unbiased feature selection?
Lukas Weber (22:52:28): > Hi all,@Stephanie Hicksand I just saw your message too – we worked on the unsupervised / clustering pipeline part of this > > Yes, if you’re interested in working together, we would definitely also be happy to help with a book. This could be super useful for people > > On the unbiased feature selection, we used SpatialDE (from Valentine Svensson), but found that it didn’t really perform any better than non-spatial HVGs in this case. So we have been actively thinking about alternatives here too. (In fact I saw something about Moran’s I statistic in another paper recently too, but can’t remember where now - seems like it might be a nice simple option) > > Also fyi in the image coordinates in your snapshot above, the coords need to be flipped and mirrored if you want to match the figures, sox = imagecolandy = -imagerow
Leonardo Collado Torres (23:05:49): > Aye, I like the idea of using the data for a new book and would be glad to contribute to it. I’ve been eavesdropping a bit lately at#osca-bookbook so I know that you are thinking on how to actually have a release and devel version, plus on a separate thing I’m going to be using GitHub actions forbookdownlater this week or the next one, so I’ll become more familiar withbookdown. > > If you want to, we could also arrange a meeting to have Kristen explain in more detail the manual annotation process using thespatialLIBDshiny app. > > As for the unbiased feature selection, I was reading a bit the wikipedia entry on Moran’s I because I’m not familiar with it. But like Lukas said, that’s an area of interest from him and Stephanie as well as us, though they are doing more work on this.
Aaron Lun (23:20:52): > The big tl;dr is that I want Bioconductor resources for analyzing spatial data with a SummarizedExperiment-class backend. Book, package, whatever, I just want to be able to point people towards it and get them off my back.
Aaron Lun (23:21:46): > If you have the basis for such a package and you want thespatulapackage name, then it’s all yours.
Stephanie Hicks (23:27:09): > hey everyone! sorry i’m late to the party. I’d be happy to help contribute to a book for spatial data withSummarizedExperiments too.@Lukas Weberjust pointed me tospatulaand I plan to look at this tomorrow to understand what all is in it. Thanks@Kevin Rue-Albrecht@Aaron Lunfor leading this effort.
Aaron Lun (23:28:35): > Ah, don’t bother with its current implementation. I realized that many of the bindings were redundant with the newersfpackage, so it’s not really necessary if you usesfinstead ofsp for representing spatial data. Right now, the main value ofspatulais in its name, which is a very cool name.
Aaron Lun (23:29:57): > The more general issue of usingsf classes for our spatial analyses is an interesting one. I doubt that anyone is doing sufficiently complex analyses to warrant the use of that level of machinery, but it’s good to try to be compatible with those classes (though avoid depending on those packages, which have complex system dependencies).
Stephanie Hicks (23:30:35): > I’m somewhat familiar withsfclasses, but definitely need to dig down into it to better understand
Aaron Lun (23:31:15): > In any case, we’re getting off point. A book and a package is what I would like to have access to by the end of this release cycle.
Aaron Lun (23:32:09): > I can commit to reviewing them, but I won’t be able to contribute more than that.
Stephanie Hicks (23:33:57): > I am on board with that and happy to contribute text & analyses.@Lukas Weber@Leonardo Collado Torreswhat do you think?
Lukas Weber (23:34:44): > Yes I would be happy to get quite involved on a book. I think this would be really useful for people, and aligns with my other work too
Stephanie Hicks (23:35:27): > my sense is@Shila Ghazanfarwould also be useful to have contributions from too
2020-05-06
Nils Eling (03:38:26): > Hi all, I guess this book would only include analysis done on transcriptomics data? Working with multiplexed imaging cytometry data, we switched to using theSingleCellExperimentclass as containers and I’d be interested to give feedback/contribute to aSummarizedExperiment-based spatial data class.
Aaron Lun (03:42:36): > From bitter experience, multiple smaller books are wise.
Nils Eling (03:44:25): > Alright, I’d like to stay in the loop to adjust the imaging cytometry workflow based on what’s developed on the transcriptomics side:slightly_smiling_face:
Aaron Lun (03:45:06): > Just start slapping a book together and I will read it.
Aaron Lun (03:45:13): > And then give my opinion.
Aaron Lun (03:45:18): > You know how it is.
Nils Eling (03:45:34): > Yeah, I missed that…
Nils Eling (03:46:07): > I think a book would be a bit much at the beginning
Aaron Lun (03:47:23): > It’s pretty easy, just take a vignette and divide it into chapters.
Aaron Lun (03:47:40): > See for examplehttps://ltla.github.io/SingleRBook/. Did it in an hour. - Attachment (ltla.github.io): Assigning cell types with SingleR > The SingleR book. Because sometimes, a vignette just isn’t enough.
Nils Eling (03:49:19): > Alright, good point. I’m guessing it’ll take me longer than an hour though:wink:
Aaron Lun (03:49:33): > Yes, probably.
Shila Ghazanfar (03:59:47): > hey i think this is really nice! My current approach for working with highly spatially resolved transcriptomics data is to use theSingleCellExperimentobject class, with each column corresponding to a cell, and include two important columns into thecolDataslot:CompressedNumericListclass objects from IRanges for the vertices of the cell segmentation mask in x- and y- dimensions. Having the cells characterised in terms of a set of vertices is quite helpful! Allows for various polygon-related geometrical operations (ie “is a point inside or outside of this polygon?“), and more relevant visualisation, I also wrote some code to either expand or contract the polygons, allowing for morphology-aware neighbourhood + subcellular analyses:thumbsup:the only catch is that your cell needs to be represented as a single (filled) polygon - no doughnuts!@Nils ElingI think the data you’re primarily working with also has a cell segmentation component, right? Would be curious to see how you’ve been tackling this > > what’s missing from this implementation, annoyingly:rolling_on_the_floor_laughing:, is that the subcellular mRNA localisation information isn’t within theSingleCellExperimentobject, but sitting in some other monster size data.frame:woman-facepalming:
Shila Ghazanfar (04:17:47): > i think it’d be important to discern the level of resolution available for various spatial technologies, it appears the use-case in mind for spatula is a single x- and y- coordinate for a cell location..
Nils Eling (05:05:52): > Yeah, I’m working more with a single-cell resolved technology. But since we only have 1um resolution, I link theSingleCellExperimentobject to segmentation masks that are stored as a list ofImageobjects:https://bioconductor.org/packages/release/bioc/html/cytomapper.htmlThat’s only for visualization for now. And I don’t need to worry about mRNA location:wink: - Attachment (Bioconductor): cytomapper > Highly multiplexed imaging cytometry acquires the single-cell expression of selected proteins in a spatially-resolved fashion. These measurements can be visualized across multiple length-scales. First, pixel-level intensities represent the spatial distributions of feature expression with highest resolution. Second, after segmentation, expression values or cell-level metadata (e.g. cell-type information) can be visualized on segmented cell areas. This package contains functions for the visualization of multiplexed read-outs and cell-level information obtained by multiplexed imaging cytometry. The main functions of this package allow 1. the visualization of pixel-level information across multiple channels and 2. the display of cell-level information (expression and/or metadata) on segmentation masks.
Lukas Weber (08:41:10): > @Shila Ghazanfarthat is very cool. Which technology are you using for the subcellular resolution?
Shila Ghazanfar (08:56:52): > thanks Lukas@Lukas Weber! im currently working with seqFISH data, but the same computational/statistical issues arise with other highly spatially resolved datasets too, like MERFISH or in situ sequencing, and to some extent to high dimensional imaging cytometry. > > it’d be good to try to think of ways of allowing generalisability when you do have subcellular or cell morphological info… after all we could frame a single point as a polygon with just one vertex:smile:or even for the visium example build a faux cell segmentation using voronoi tesselation ?
Shila Ghazanfar (09:00:12): > i’m also keenly aware that the technology is moving fast - we should be anticipating being able to resolve 3D cellular structures, and representing these perhaps as solid irregular polyhedra:face_with_monocle:
Sean Davis (09:24:19): - Attachment: Attachment > <!channel> reminder that although there are volunteers for the following several weeks of single-cell multimodal journal club, there’s still no presenter for next week May 11. I think we’re hesitant to start random assignments on this short notice, so if there’s no volunteer we’ll have to skip a week. https://github.com/waldronlab/data-science-seminar/wiki/Single-cell-multimodal-data
Dario Righelli (11:25:08): > @Aaron LunI saw your spatula package and I think we can overlap with some features with ourSpatialExperimentandVisiumExperimentclasses, inheriting fromSingleCellExperiment. > Maybe we can find a way to let them work together. > (https://github.com/drighelli/SpatialExperiment)
Dario Righelli (11:26:06): > Here is an example of analyses for both classes.https://github.com/drighelli/SpatialAnalysis
Lukas Weber (11:48:10): > @Shila Ghazanfaryes we are mainly working with Visium, so I haven’t thought much yet about how to include subcellular resolution. I did just read the latest MERFISH paper in some detail - which I think is very similar to seqFISH. I think the idea of putting IRangesLists into colData is a great idea > > Yes I wonder if the single-cell-level or spot-level data (e.g. Visium) could end up being a special case of a more generalized class, or if it ends up simpler to have multiple classes, not sure - pretty sure we had an earlier discussion along these lines a while back too > > And yes definitely 3D will be coming along at some point too, which will make things even more complicated:joy:
Hervé Pagès (11:50:36): > @Shila GhazanfarIt’s interesting that you’ve managed to use an IRangesList object to represent a collection of polygons. Did you introduce an artificial offset on the y coordinates of the vertices to work around the constraint thatend(your y coordinates) must be >=start-1(you x coordinates) in an IRanges object?
Shila Ghazanfar (11:58:33) (in thread): > ah, apologies! the correct class I’ve used isCompressedNumericListstill defined from IRanges package, which as far as my experience goes, doesn’t have these constraints you mentioned
Hervé Pagès (12:43:19) (in thread): > I see. Excellent! and cleaner than hijacking IRanges objects for representing points in the 2D space. Sounds like we should have a class for this though so you don’t need to use 2 colData columns for that. Maybe something like a SplitMatrixList that would be general enough to support the 3D usecase.
Shila Ghazanfar (12:49:10) (in thread): > yes! a SplitMatrixList is definitely a more elegant & generalisable solution than the hodge-podge I have going on right now:smile:
Hervé Pagès (13:06:02) (in thread): > … even though the 3D case will need some extra slots to group the vertices in edges and faces. mmh.. that’ll be interesting to explore.
Hervé Pagès (13:07:16) (in thread): > unless the shapes are convex, in which case the faces and edges are implicit
Shila Ghazanfar (13:53:16) (in thread): > Yeah, i think we might have to stick with the convex assumption in the 3D case, either way it’s interesting and worthwhile to be able to generalise… > > another option is to consider a list of N_z polygons, representing the cross-sectional polygons when you cut through each of the N_z z-planes. I think for a fair number of technologies there is high resolution in x- and y-, but the z-dimension is quite distinct and maybe corresponds to just tens of integer values
Hervé Pagès (14:08:07) (in thread): > Oh I see. That would actually keep things relatively simpler, especially if, as you say, the number of z-planes is small and the planes are the same for everybody.
Aaron Lun (14:21:05): > If I may throw in my 2c: I’d be wary of homebrewing our own low-level classes here. There’s an established ecosystem of R packages devoted to spatial analyses, so it seems best to plug into that where we can. For example, thespclasses are quite old (2005; almost as old as BioC itself!) and have hundreds of downstream dependencies. While those particular classes have quite a few flaws, thesfclasses seem pretty functional and you can directly stuff them into thecolDataof a SE without any problems. Unless proven otherwise, I would say that this is sufficient for most of our use cases.
Sean Davis (14:49:40): > I would second looking at the sf package. I’ve been using it extensively for non-single-cell datasets and it has a large number of methods, good documentation, and plays nicely with both base R and s4 classes.
Hervé Pagès (14:50:21): > I take this as a hint for me to seriously start looking at spatula. Just started doing this a few minutes ago. Oh thespclasses are actually S4 classes and have a vector-like semantic! That is indeed great news. I like what spatula is doing with that, implementing methods for core low-level things likeextractROWS()andbindROWS(). mmm yummy! I’m proud of you guys:smiley:
Hervé Pagès (15:10:45): > In spatula’s vignette: > > Sorting is performed based on a comparison of the vector of coordinates for the cluster centroids; if those are equal, then polygons are sorted by area. > uh, nice! > What are you guys waiting for submitting this?
Hervé Pagès (15:17:19): > Now I can useexpandPoints(points, radius=1)to make bubbles:wink:
Aaron Lun (15:37:25): > Yes, I thought you might like it.
Aaron Lun (15:39:08): > The big question is whether to usesforspclasses. The former are superior in every respect; more coherent, easier to use, directly compatible withbindROWSand friends (so we don’t have to write our own things to handle them). The only disadvantage is their use of some hard-core system dependencies (GEOS, GDAL and PROJ) thatspdoes not need.
Aaron Lun (15:40:12): > If we decide to go withsfclasses, then many of the utilities inspatulacan be discarded, possibly up to the point that we wouldn’t even needspatula at all.
Aaron Lun (15:41:05): > It would have been nice if thesfauthors split the class definition from the methods that rely on those system dependencies, but I can sort of understand why they did it like this, given that the classes are mostly useless without the methods.
Sean Davis (15:47:55): > I looked at both sp and sf. The sf functionality is cleaner and more complete, in my limited experience.
Hervé Pagès (15:52:56): > but a sfatula package wouldn’t be so exciting
Aaron Lun (15:53:35): > I won’t say that wasn’t a consideration.
Hervé Pagès (16:38:48): > Aren’t some of the operations implemented on top of GEOS, GDAL and PROJ actually things we’d also like to be able to use? Or we only care about the classes defined insf?
Aaron Lun (16:40:17): > Yes, they are, which is why I understand why they depended on it. The classes would have very little utility without at least GEOS, for example.
Hervé Pagès (16:44:06): > So we are kind of in the same situation. We want thesfclasses for an enhanced SingleCellExperiment container but we also want thesfmethods for some of the operations we’re going to support on this enhanced SCE. I guess I don’t really understand how it would have helped if thesffolks had separated the methods from the classes so I kind of feel that I’m missing something.
Hervé Pagès (16:48:50): > Or maybe you’re saying that we only need GEOS so it would’ve been better if they only depended on that.
Hervé Pagès (16:52:48): > Anyway, I just want to emphasize again the fact that installingsfis not a problem on Windows or Mac because CRAN provides statically linked binaries for these packages. That’s when the binaries are available of course but AFAIK that’s always been the case in release so at least the end-user is covered. People using devel are on their own, as always. It’s also pretty easy to install them on Linux (just need a few additionalapt-get installiterations).
Aaron Lun (17:38:08): > The main difference is that if I had an SCE with sf objects embedded within, and I wanted to load the SCE object but didn’t care about the sf objects, I wouldn’t want to drag in GEOS, etc. that I wasn’t planning on using.
Aaron Lun (17:40:26): > In this case it’s probably fine as those sf objects will collapse down to data.frames anyway (I think).
Hervé Pagès (19:58:49): > They are S3 classes, so, unlike S4 classes, there is no formal class definition. If you load an SCE withsfobjects embedded within, and you don’t care about thesfobjects, then there is no need to load thesfpackage. And you can still do the usual basic manipulations on the SCE like[,cbind, etc… and the embeddedsfobjects will follow. Because, yes, as you said, they are basically data.frame objects (with some extra attributes): > > > class(nc) > [1] "sf" "data.frame" >
2020-05-08
Aaron Lun (13:33:52) (in thread): > I realized I never responded to this. Maybe checkout our discussion below aboutsfclasses, and whether that might be a good way to try storing the coordinate data.
Aaron Lun (13:34:30) (in thread): > Certainly it would make life easier for people doing routine geometric operations, e.g., which points overlap my tumor region.
2020-05-19
Aedin Culhane (01:07:50): > Has anyone looked at spDataLarge fromhttps://nowosad.github.io/drat/
Tobias Hoch (11:16:45): > @Tobias Hoch has joined the channel
2020-05-21
Kylie Bemis (16:17:28): > @Kylie Bemis has joined the channel
2020-05-22
Aaron Lun (00:15:37): > okay. it’s been 2 weeks. Where’s the book?
Lukas Weber (00:36:54): > haha yes good reminder!:joy:yes me and@Stephanie Hickswere talking some more about a book. Will get something started and post some more here soon:+1:
2020-06-03
Chris Vanderaa (04:57:44): > @Chris Vanderaa has left the channel
2020-06-06
Olagunju Abdulrahman (19:58:15): > @Olagunju Abdulrahman has joined the channel
2020-06-09
Aaron Lun (15:17:15): > BOOK.
Stephanie Hicks (19:18:14): > So@Lukas Weberhas actually started reading about rebook!
Aaron Lun (19:18:30): > what’s rebook?
Aaron Lun (19:18:33): > Oh, my thing.
Aaron Lun (19:18:38): > is that my thing?
Stephanie Hicks (19:18:53): > Yes
Stephanie Hicks (19:19:11): > Trying to figure out how to set up the book
Aaron Lun (19:19:37): > You could probably just grab OSCA repo and throw away all the Rmd’s and that’s your set up right there.
Stephanie Hicks (19:19:54): > Oh ok. That works too.
Aaron Lun (19:20:16): > Ovbviously have to change all the “rochestrating tblah blah blah” text.
Stephanie Hicks (20:31:13): > Ok we will start on that then
2020-06-11
Aedin Culhane (03:05:10): > I haven’t verified on sc data, but in the vignette spdep claim it is faster that sf for nb operations such as poly2nb. Seehttps://cran.r-project.org/web/packages/spdep/vignettes/nb_sf.htmlHas anyone looked at this, or checked if the sfnetworks package addressed polygon neighbors?
Stephanie Hicks (23:19:37): > i haven’t
2020-06-12
Aedin Culhane (13:20:16): > Thanks stephanie
Shila Ghazanfar (15:16:45): > im not sure how relevant this is given the other spatial packages, i’ve written some (very slow:grimacing:…) code to take the vertices of polygons and build up a neighbourhood network, with an expansion tuning parameter, inhttps://github.com/shazanfar/spatialStuffin the reports folder.. the nice thing is that the spatial coords dont necessarily need to be pushed into a special class, just data.frame - File (PNG): image.png
2020-06-16
Paul Hoffman (18:50:26): > @Paul Hoffman has joined the channel
2020-06-17
Ruben Dries (12:29:25): > @Ruben Dries has joined the channel
2020-06-19
Huipeng Li (13:25:38): > @Huipeng Li has joined the channel
2020-06-25
CristinaChe (15:59:26): > @CristinaChe has joined the channel
2020-06-27
Lukas Weber (00:35:10): > Hi all, ok here is a repo with a start at a spatial book, as promised a while back:tada:https://github.com/lmweber/OSTA-base. It is a work in progress - but the general outline and a fair bit of initial content is there now, so I think we can start thinking about collecting input from others who are interested! > > So far, it is focused on Visium, using our human brainspatialLIBDdataset (from@Leonardo Collado Torres) for the examples. (There is a main workflow inhuman_DLPFC.Rmd, and objects from there are used in the earlier analysis chapters.) But one of the things we would really like to hear from others about is other technologies to include - e.g. the work@Shila Ghazanfarhas been doing with seqFISH. > > I have set it up using@Aaron Lun’s setup from OSCA, i.e.baseandreleaserepos, and usingrebookto link objects. Right now it is all in my GitHub, although I haven’t pushedhtmlfiles yet since they keep changing. If you want to have a look at it, probably easiest to clone it and compile it locally - you need a few dependencies likescater/scranplus my own package-in-progresshttps://github.com/lmweber/spatzliwhere I have been collecting a bunch of functions. It compiles in a few minutes. Let me know if this does/doesn’t work or if I should do this differently. > > Right now, there are a few specific things that I wanted to ask people about: > > - data structure: I usedSingleCellExperimentsfor now, but we would like to update this to@Dario Righelli’sSpatialExperiment, which looks very useful especially with thespatialCoordsslot. Also I believe@Shila Ghazanfarhas some object structures for working with polygons/centroids (instead of only Visium spots), which would also be very cool to integrate somehow. Any suggestions/ideas/comments here? (or other planned extensions toSpatialExperiment?) > > - technologies: to add new technologies/datasets, I think a good way to start is to add another workflow (either short, or longer with lots of detail), and then we can think about how to also integrate some of this into the earlier chapters usingrebook.@Shila Ghazanfarwould you be interested in starting a seqFISH workflow (or any other ideas here?) > > - datasets: I have also started a package to collect the example datasets athttps://github.com/lmweber/STdata, which will turn into an ExperimentHub package. Any other suggestions here? > > Then over the longer term, as this starts turning into a more complete pipeline,@Stephanie Hicksand I are thinking we will also aim to write this up in some sort of OSTA paper (similar to the OSCA paper), with co-authorships from everyone here who is interested in contributing. For now, I will keep adding additional material (currently working on the feature selection chapter) - let me know if anyone is interested in adding material in any particular parts, or any other feedback!
Aaron Lun (16:57:42): > YES
2020-06-28
Aaron Lun (04:40:07): > I look forward to a draft release of some content.
Dario Righelli (07:53:39) (in thread): > Great work! Thanks for mentioning my packageSpatialExperimentwhich of course needs some more work to be done (any suggestion is appreciated). > For the ExperimentHub package, I contributed to theSingleCellMultiModalpackage of@Marcel Ramos Pérezby adding the seqFISH dataset combined with a scRNAseq dataset produced for the BIRSBIOIntegration Hackathon (leaded by@Aedin Culhane@Kim-Anh Lê Caoand@Elana Fertig). > Note that for now the seqFISH data are available only on my fork (https://github.com/drighelli/SingleCellMultiModal), because it uses theSpatialExperimentpackage not yet available on Bioconductor. > Maybe it could be useful to add more spatial datasets on the Marcel’s package.
Lukas Weber (20:08:13) (in thread): > Thanks! I will have a look at theSingleCellMultiModalpackage to see how you have set it up there. I feel it will probably end up being simpler to have a separate package with just the datasets for this book, since this will make it easier to maintain things all in one place. But I will think about this some more - and seeing an example of how you have already usedSpatialExperimentis also helpful, thanks!
2020-06-29
Dario Righelli (03:05:12) (in thread): > If you look at the vignette for the seqFISH data, you’ll see that the dataset is stored into aMultiAssayExperimentobject where you can find the seqFISH data as aSpatialExperimentobject and the scRNAseq data as aSingleCellExperimentobject. > The good part is that they both can be easily stored into a MAE object.:slightly_smiling_face:
Lukas Weber (08:34:15) (in thread): > ok thanks!
2020-07-02
Aaron Lun (20:17:16): > I suppose I should polish uprebookand get it onto the Bioc build system.
2020-07-04
Umar Ahmad (08:21:31): > @Umar Ahmad has joined the channel
2020-07-07
Dario Righelli (11:21:43): > Hi<!here>, I wanted to inform you that I just submitted theSpatialExperimentto Bioconductor.:star-struck:
Aaron Lun (11:21:59): > Allright
Lukas Weber (11:22:03): > awesome, nice work
Martin Morgan (12:59:46): > @Dario Righelliwe’re piloting a new workflow for reviewing packages – they get added togit.biocondcutor.orgfirst, then reviewed and if accepted added to the manifest, instead of reviewing, add togit.bioconductor.org, add to manifest. Can we use your package as a test case? > > The updated contributions instructions are athttps://github.com/Bioconductor/Contributions/tree/git-first-readmebut essentially the change is that you start pushing togit.bioconductor.orgimmediately after your package enters the review process… > > There’s also an update set of instructions for pushing togit.bioconductor.orgathttps://github.com/Bioconductor/bioconductor.org/blob/nturaga-master/content/developers/how-to/git/new-package-workflow.mdAs a beta user your feed back on the process, including the instructions, would be great!
Dario Righelli (17:24:50) (in thread): > Yeah sure I can do it! I’ll try to upload the package to thegit.biocondcutor.org.
Martin Morgan (17:43:04) (in thread): > Actually, that’s part of the process; we’ll do that…
2020-07-14
Celso Milne (19:16:47): > @Celso Milne has joined the channel
2020-07-16
Jayaram Kancherla (09:44:07): > @Jayaram Kancherla has joined the channel
2020-07-23
Charlotte Soneson (07:28:46): > “STUtility is an R-package with the goal of providing an easy-to-use visualization and analysis tool kit for spatial transcriptomics data.”https://ludvigla.github.io/STUtility_web_site/index.html(described inhttps://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-020-06832-3) - Attachment (BMC Genomics): Seamless integration of image and molecular analysis for spatial transcriptomics workflows > Recent advancements in in situ gene expression technologies constitute a new and rapidly evolving field of transcriptomics. With the recent launch of the 10x Genomics Visium platform, such methods have started to become widely adopted. The experimental protocol is conducted on individual tissue sections collected from a larger tissue sample. The two-dimensional nature of this data requires multiple consecutive sections to be collected from the sample in order to construct a comprehensive three-dimensional map of the tissue. However, there is currently no software available that lets the user process the images, align stacked experiments, and finally visualize them together in 3D to create a holistic view of the tissue. We have developed an R package named STUtility that takes 10x Genomics Visium data as input and provides features to perform standardized data transformations, alignment of multiple tissue sections, regional annotation, and visualizations of the combined data in a 3D model framework. STUtility lets the user process, analyze and visualize multiple samples of spatially resolved RNA sequencing and image data from the 10x Genomics Visium platform. The package builds on the Seurat framework and uses familiar APIs and well-proven analysis methods. An introduction to the software package is available at https://ludvigla.github.io/STUtility_web_site/ .
Aaron Lun (11:15:31): > Hm. Dependency on NNLM.
Lukas Weber (12:17:12): > what are issues with NNLM? google says it has been removed from CRAN
Hervé Pagès (12:44:26): > https://cran.r-project.org/web/packages/NNLM/index.html
Aaron Lun (12:45:06): > Putting that aside, the book hasn’t seen a commit in a month. What’s the deal?
Lukas Weber (13:01:18): > yep been working on some other projects too. will get back to it again soon
Stephanie Hicks (14:51:55): > ha, trying to install now to give a test run. found a minor typo that made me laugh:upside_down_face: > > devtools::install_hithub("linxihui/NNLM") >
Lukas Weber (14:54:34): > looks like it was dropped from CRAN for failing checks. But the Travis build is still passing
hcorrada (15:01:31) (in thread): > register that name if nobody has it! you can get going on that music production business
Stephanie Hicks (15:15:59): > It’s going on 15+ mins for installation…
Casey Greene (16:03:44) (in thread): > I have written GitHug into a grant before
Casey Greene (16:04:20) (in thread): > One of those typos found before submission fortunately
2020-07-25
Dr Awala Fortune O. (16:29:45): > @Dr Awala Fortune O. has joined the channel
2020-07-29
Spencer Nystrom (11:48:15): > @Spencer Nystrom has joined the channel
Leonardo Collado Torres (15:56:28): > Hi! So, I guess@Dario Righelli’s packagehttps://github.com/drighelli/SpatialExperiment/will be the main one moving forward for the data structures for this type of data, right? I’ll need to look closer into it beyond the questions I asked during his presentation today. It looked quite complete though and congrats on the excellent work!
Leonardo Collado Torres (15:57:28): > For future ref, these were my questions: > > * Dario, how would you deal with different quality versions of the histology images? In our case we have low, medium and high (about 500 mb each) images but in the future we might have super high (several dozen GB per image). Should the images still be a part of the object or links to say AWS? Also, say you have an image mask, where do you store it? > * Dario, do you see any benefit on having altExp()s for the spot level data and say pseudo-bulked layer-level data? Or would you keep them as separate objects? >
Ellis Patrick (15:58:53): > I’m guilty for this one… Dario - Can you see SpatialExperiment also holding morphology information, cell masks or even images?….
Ellis Patrick (15:59:03): > I’d love to hear more conversation around this
Dario Righelli (15:59:14): > Thanks@Leonardo Collado Torres!!! We have for sure continue to discuss about your points!:slightly_smiling_face:
Ellis Patrick (15:59:15): > it would be worth looking at cytomapper too
Dario Righelli (16:00:01): > for the morphology information I suppose that you can add column(s) to the spatialCoords(spe) as you prefer
Ellis Patrick (16:00:50): > I think it might make sense to have another assay?
Dario Righelli (16:01:13): > I saw that the Giotto tool uses the morphology starting from the SingleCellExperiment
Aaron Lun (16:01:28): > It could fit in as an altExp or a colData element, depending on how you like to take it.
Dario Righelli (16:02:20): > indeed, as@Aaron Lunsuggested for our package, we stored the spatial coordinates as part of the int_colData
Shila Ghazanfar (16:02:50): > for the cell morphology, I store the vertices of the cell boundaries as anIRanges::CompressedNumericListin the colData() of an SCE object, one for x- and y-. it’s obviously making do with an existing structure intended for another purpose & does the job so far:thumbsup:but perhaps will be more generalised, especially to keep x- and y- linked with each other
Dario Righelli (16:03:26): > @Shila Ghazanfaryes, and I just put those structures into the spatialCoords of our object
Ellis Patrick (16:03:54): > cell profiler spits out around 20 morphology measurements so not an insane amount to store
Dario Righelli (16:04:28) (in thread): > the ones interested in Shila’s work can take a look athttps://github.com/shazanfar/spatialStuff/blob/master/reports/SpatialExperiment_polygons_example.html
Dario Righelli (16:07:06) (in thread): > Thanks I already kept notes of them!:wink:
Dario Righelli (16:07:45): > One of the things that I was wondering is the image storing
Ellis Patrick (16:07:49): > however…. altExp sounds like it would be more appropriate for my type of data…. then you could store mean instensities, median intensties etc
Dario Righelli (16:08:16): > it sounds like a huge amount of data to keep in memory, in particular when you have a lot of them
Shila Ghazanfar (16:08:55): > just thinking for seqFISH data, where you also have the underlying mRNA spots (subcellular level), would altExp be useful for storing the mRNA spots in a cohesive way
Dario Righelli (16:09:23) (in thread): > mmm can I take a look at your code?
Ellis Patrick (16:09:34): > probably with spots is whether you want to assign/map them to cells or not….
Shila Ghazanfar (16:10:13) (in thread): > yes that’s true, to be fair I think it’s better to retain all mRNAs detected in the image, even if they fall outside of segmented “cells”
Ellis Patrick (16:10:14): > Does that start to play into multiassay territory? Where not all spots will be in a cell
Dario Righelli (16:12:19): > If you are referring to the MultiAssayExperiment it wants an overlap between multiple experiments
Ellis Patrick (16:13:28) (in thread): > This makes no sense at all…. because the spot is the feature….
Ellis Patrick (16:15:57) (in thread): > @Shila GhazanfarI can’t even imagine how this would look… you effectively just have 1000000 spots, each with a x, y, RNAlabel, cellLabel ?
Shila Ghazanfar (16:17:25) (in thread): > each spot can be thought of an observation, so i would imagine this akin to storing each read in high throughput sequencing. and yes currently i have it as one ginormous data.frame:exploding_head:
Ellis Patrick (16:19:45) (in thread): > an n x 4 data frame or 4 x n ? Do people whinge about this a lot? I still can’t handle my SingleCellExperiment having 50 rows and 100000000 columns
Shila Ghazanfar (16:22:48) (in thread): > basically an nx4data.frame, it’s not that often that I call it, but extremely useful to have
Ellis Patrick (16:26:04) (in thread): > do people whinge about p x n here very much? It made sense in microarray world, but I’m wondering if we’ve progressed way passed that now that we are measuring millions of cells…
Riyue Sunny Bao (19:20:25): > @Riyue Sunny Bao has joined the channel
Brianna Barry (21:34:28): > @Brianna Barry has joined the channel
2020-08-05
Hans-Rudolf Hotz (03:23:32): > @Hans-Rudolf Hotz has joined the channel
shr19818 (13:46:22): > @shr19818 has joined the channel
2020-08-13
Loyal (14:25:06): > @Loyal has joined the channel
2020-08-18
Will Macnair (09:08:32): > @Will Macnair has joined the channel
Leonardo Collado Torres (11:22:23): > Hi, I was checking outhttps://events.linuxfoundation.org/r-medicine/program/schedule/($10 USD for students) and noticed: > > {nDSPA}: An R/Bioconductor Package for Quality Metrics, Preprocessing, Visualization, and Differential Testing Analysis of Spatial Omics Data > Riyue Bao > > However,http://bioconductor.org/packages/nDSPAdoesn’t exist. Are any of you familiar with this package? It might have another name as I understand that sometimes people submit abstracts before they have a finished product. - Attachment (Linux Foundation Events): Schedule | Linux Foundation Events > Thursday, August 27 | 11:00 AM – 7:00 PM EDTShort Courses* 11:00 – 15:00101: Intro to R for Clinicians Stephan Kadauke, Amrom Obstfeld, Joe Rudolf15:00 – 19:00201: Intro to Machine Learning with…
Lukas Weber (11:40:28): > nope, not familiar with this one
Hervé Pagès (12:16:54): > doesn’t seem to be on CRAN or GitHub either
Lukas Weber (12:18:27): > I had a quick search around too, couldn’t find anything
Simina Boca (12:55:46): > I found an ISCB abstract that says “nDSPA will be released on GitHub and accompanied by an R Shiny application for interactive data exploration”https://www.iscb.org/cms_addon/conferences/ismb2020/posters.php?track=HitSeq%20COSI&session=B - Attachment (iscb.org): Home > ISCB - International Society for Computational Biology
Simina Boca (12:55:58): > That abstract also says it’s an R/Bioconductor package though
Simina Boca (12:56:14): > Not really OK to say that if it’s not in Bioconductor - in this case it’s not even on GitHub!
Alan O’C (13:03:05): > Seems bizarre to present a software package without at least putting it on github
Aaron Lun (13:03:43): > IP, obviously
Hervé Pagès (14:12:20): > The package is most likely hiding in a private repo somewhere herehttps://github.com/riyuebao
Alan O’C (14:35:43) (in thread): > I don’t follow
Aaron Lun (14:35:58) (in thread): > intellectual property
Alan O’C (14:36:27) (in thread): > I know the term but not why it would be a concern here
Aaron Lun (14:37:22) (in thread): > should have put a /s
Simina Boca (14:45:00) (in thread): > But if they want it on BioC, what IP?
Alan O’C (14:45:49) (in thread): > I think we’ve killed Aaron’s joke and are now parading around on its grave
Alan O’C (14:47:31) (in thread): > Anyway presenting scientific software without making it available is basically just gloating
Simina Boca (14:52:03) (in thread): > In this case saying that it’s a Bioconductor package is also… lying?
Alan O’C (14:52:39) (in thread): > Yes, lying and using the good reputation of Bioc packages as (false) advertisement
2020-08-19
Riyue Sunny Bao (15:00:26): > Hi!! thanks for the comments! we took it off as we found some bugs in the stats module. so we are revising it and will release it before R/med!
Riyue Sunny Bao (15:03:06): > it is also not on bioc, as we are running behind on our submission schedule. I will contact the organizer to change the title to R and remove bioc if that makes sense
Riyue Sunny Bao (15:03:09): > sorry about that!
Riyue Sunny Bao (15:37:16) (in thread): > Not the IP:sweat_smile:we did not file any IP on this, and did not make sense since bioc has its own policy in IP protection. But I guess I get the joke….
Hervé Pagès (15:58:42): > No problem. Thanks for the update. Looking forward for your submission to Bioconductor.
Aaron Lun (16:14:34) (in thread): > you guys are lucky, I’ve always got to think about IP. Gotta remember to avoid accidentallygit commiting all those cancer cures.
Riyue Sunny Bao (16:19:31): > yup will do - thehow to make a bioc packagetalk by Kayla Interdonato at bioc2020 was super informative:slightly_smiling_face:
2020-09-11
Matt Stone (12:49:00): > @Matt Stone has joined the channel
2020-09-23
Giuseppe D’Agostino (22:05:52): > @Giuseppe D’Agostino has joined the channel
2020-10-08
Davide Corso (05:29:50): > @Davide Corso has joined the channel
2020-11-10
Stephanie Hicks (09:32:28): > for anyone working in inferring cell-to-cell communication using some type of#spatialgene expression data, this seemed like a useful reviewhttps://www.nature.com/articles/s41576-020-00292-x - Attachment (Nature Reviews Genetics): Deciphering cell–cell interactions and communication from gene expression > Cell–cell interactions and communication can be inferred from RNA sequencing data of, for example, ligand–receptor pairs. The authors review insights gained and the methods and tools used in studies of cell–cell interactions based on transcriptomic data.
2020-11-15
Shila Ghazanfar (11:19:59): > Hey everyone, quick question re performing batch correction (ideally with MNN). Say I have a large heterogeneous set of data with batches. I perform PCA and then fastMNN, and now I have a set of corrected PC values, corresponding to genes that are most highly variable among the large heterogeneous set. > > Now I wish to subset down to a specific type within the data, and this type is quite likely confounded with the batch^. Indeed if I subset to this data, perform PCA and then fastMNN, I’m left with an extremely homogeneous corrected data (and have removed biology due to the confounding). However, if I simply extract the corrected PC values from the initial batch correction above, the biological signal corresponding to the specific type is buried much lower among the later PCs, meaning whatever downstream analysis I do (clustering, trajectory, etc) is noisy. > > And so, apart from the option of just not performing any batch correction, I’m wondering if there exists some approach to re-weight or re-calculate the PCs given the batch correction association that’s performed with all the heterogeneous data? Unsure if I’ve missed some key feature somewhere, where you can port the correction vectors over just for a specific type within the larger data context. I hope that makes sense and is something that others may have come across..? thanks :) > > ^this is not super likely to happen for scRNA-seq, but when batch is confounded with spatial region, and biology is spatially regionalised, then..:shrug:
Aaron Lun (15:30:21): > If you already managed to subset your data, then it seems like you’ve already found some way of deciding that the subset of cells is somehow “the same” across your samples.
Aaron Lun (15:31:14): > So the homogeneity introduced by the second round of correction seems entirely appropriate to me.
Aaron Lun (15:32:43): > I would question why you even want to do the second round of correction if you’ve already managed to identify the cell type that you care about.
Shila Ghazanfar (16:08:50): > it comes down to a question of the degree of technical effect. I know that the subset is not entirely the same(homogeneous), and so the extremely homogeneous corrected data is a red flag. But i don’t know the extent of technical/batch effect that may persist alongside if i don’t explicitly correct. the aim in this case is to enable downstream analysis for the subset e.g. clustering
Shila Ghazanfar (16:13:19): > is it that there isn’t really an existing way to re-weight a subset given the ‘batch correction vectors’ or something along those lines of the superset of cells? just want to know in case ive missed some approach I could be employing in this case :)
2020-11-30
Sean Davis (11:19:00): > Title:Examination of cell-lineage and inter-cell interactions from spatial transcriptomics dataAbstract: The spatial transcriptomics (ST) technology has enabled geographical profiling of tumor gene expression. However, each ST spot may detect mixture signals from diverse immune or malignant cells of unknown lineages, and local tissue densities may vary significantly across regions. Therefore, the decomposition of ST cell lineages remains a challenge that cannot be resolved by previous decomposition methods for fixed cell types in bulk tumors. We developed the Spatial Cell Estimator (SpaCE) to infer the cell identities and intercellular interactions for tumor ST data. Based on reliable cell lineage inference, SpaCE can further reveal how intercellular interactions affect the pathway and gene activities in distinct regions to modulate the cancer progression.Bio: Beibei Ru is a postdoctoral research fellow in Dr. Peng Jiang’s Lab at NCI/CDSL. He is developing tools to mining spatial transcriptomics data. Prior to joining Peng’s Lab, Beibei did his PhD at the University of Hong Kong where he investigated the aberrant epigenetic regulation in cancer development.Meeting details:Time: Nov 30, 2020 03:00 PM Eastern Time (US and Canada) > > Please download and import the following iCalendar (.ics) files to your calendar system. > Weekly:https://umd.zoom.us/meeting/tJUlc-qoqTorGNEFjhGcqXTkFz0DZS2IKlGq/ics?icsToken=98tyKuCppj8pGtOUsRuCRowcGo-gb_Twtn5cjY14uhXtJCNCWjf9EPgSFohMQvH7Join Zoom Meetinghttps://umd.zoom.us/j/91843071125Meeting ID: 918 4307 1125 > One tap mobile+13017158592,,91843071125#US (Washington D.C)+19294362866,,91843071125#US (New York) > > Dial by your location > +1 301 715 8592 US (Washington D.C) > +1 929 436 2866 US (New York) > +1 312 626 6799 US (Chicago) > +1 346 248 7799 US (Houston) > +1 669 900 6833 US (San Jose) > +1 253 215 8782 US (Tacoma) > Meeting ID: 918 4307 1125 > Find your local number:https://umd.zoom.us/u/aHtAdmaOT
2020-12-01
Aedin Culhane (13:53:27) (in thread): > Hi Shila Do you just want to remove or downweight specific PCs. This is easy to by setting the eigen value to 0 or reducing it, you can they back multiple the 3 matrices and redo the PCA. We implemented this in mogsa, (for multi modal data)
Aedin Culhane (13:56:00) (in thread): > You can also project a matrix onto an existing PCA, to do so you need either matching rows or features, However you can iteratively align and input if you have missing objects. See suprow/supcol, suppl in in ade4, made4 and in corral which is better for scRNAseq.
Aedin Culhane (13:56:47) (in thread): > Happy to video chat and explain in more detail if you wish
2020-12-12
Huipeng Li (00:38:14): > @Huipeng Li has joined the channel
2020-12-13
Kelly Eckenrode (13:42:16): > @Kelly Eckenrode has joined the channel
2020-12-14
Nick Owen (13:21:53): > @Nick Owen has joined the channel
FelixErnst (14:01:43): > @FelixErnst has joined the channel
Alsu Missarova (14:07:29): > @Alsu Missarova has joined the channel
2020-12-15
Alsu Missarova (10:21:40): > Following up on my question in bioc2020 conference yesterday (re storing graph-based data (e.g. cell-to-cell neighbourhoods)) - it is possible to add an igraph slot on top of SingleCellExperiment object. For example, it has be done (for different purposes) for the package Milo for identifying DA between scRNA-seq datasets -https://github.com/MarioniLab/miloR
Ludwig Geistlinger (10:30:34) (in thread): > Sounds like this might be relevant for you:https://www.bioconductor.org/packages/release/bioc/html/TreeSummarizedExperiment.html - Attachment (Bioconductor): TreeSummarizedExperiment > TreeSummarizedExperiment has extended SingleCellExperiment to include hierarchical information on the rows or columns of the rectangular data.
Ludwig Geistlinger (10:30:58) (in thread): > see also#tree-like-se
Dario Righelli (10:46:15) (in thread): > Hi@Alsu Missarovathanks for asking again this. As we mentioned yesterday during the session, I think that it could be useful to add a graph managing slot or something similar to the class. > We are constantly developing the class to leave it as light as possible (in terms of dependencies) and to give something usable for everyone. > Btw we’re putting a lot of thoughts into this class and I’m sure we’ll come out with a final version including something for graphs (it could be theiGraphin the end ^^’) in a short hand…
Dario Righelli (10:46:42) (in thread): > I’m going to add this point to our spatial challenges, because it’s a really important point!
Dario Righelli (10:48:47) (in thread): > Btw@Ludwig GeistlingerI think that in the end also theTreeSummarizedExperimentwill be useful for the spatial data. > I still don’t know if to inherits from that one or not ^^’
Aaron Lun (11:27:55) (in thread): > That’s what thecolPairs()is for.
Dario Righelli (11:44:52) (in thread): > So we already have a solution!:smile:
Aaron Lun (11:46:10) (in thread): > Though do read the warnings in the documentation about the inconsistency of subsetting with a pair-like structure.
Dario Righelli (11:49:23) (in thread): > Just thinking about the spatial organisation of the “cells” into the graph/pairs, but I think it’s up to the applied method
2020-12-18
Milan Malfait (09:42:58): > @Milan Malfait has joined the channel
2020-12-21
Dario Righelli (15:03:19) (in thread): > I’m sorry to be so late, but I just realized that I never reported any feedback, but anything went really good on my side! ^^ > Sorry about that!
Martin Morgan (15:15:39) (in thread): > thanks, no problem!
2021-01-01
Bernd (14:07:03): > @Bernd has joined the channel
2021-01-22
Annajiat Alim Rasel (15:45:37): > @Annajiat Alim Rasel has joined the channel
2021-02-23
Wynn Cheung (10:33:09): > @Wynn Cheung has joined the channel
2021-03-20
watanabe_st (01:58:31): > @watanabe_st has joined the channel
2021-03-23
Lambda Moses (23:06:13): > @Lambda Moses has joined the channel
2021-04-08
Jenny Drnevich (00:22:28): > Has anyone analyzed CARTANA spatial data in R? I just got a project dropped in my lap with the data processed with CARTANA ISS service and all I have is a bunch of image directories and a two-page pdf with instructions like “To view images, you will need MATLAB installed.” and the original publication has “All code was written in MATLAB and is freely available athttps://github.com/kdharris101/iss”:sob:. 10X did just acquire CARTANA but they do not have any software yet. I would greatly appreciate any suggestions or better yet, pointers to R codes!
Lukas Weber (09:22:46) (in thread): > @Helena L. Crowell@Mark Robinsonpossibly might know some more about this
Helena L. Crowell (09:25:11) (in thread): > Yes, well, to be honest: We received.rdsobjects from prior preprocessing in MATLAB (by someone else) - following the pipeline suggested by CARTANA at the time I believe. Not sure how one would go about this in R. Then again, things might change once 10X gets things up and running.
Jenny Drnevich (10:47:56) (in thread): > Thanks@Lukas Weberand@Helena L. Crowell. I will contact 10X but my guess is that they will take a while to get something out. I’ll find local MATLAB expertise and see if I can get someone to do the pre-processing for me. But the base data must be multiple fluorescence intensities per x-y coordinate, yes?
Mark Robinson (11:11:00) (in thread): > @Jenny Drnevichdoes CARTANA not deliver other files other than the images? .. one of the previous datasets arrived with something like this screenshot .. so it had the images and the MATLAB code, but also some CSV outputs. If you don’t have that, then yeah, hopefully someone local can help. - File (PNG): Screenshot 2021-04-08 at 17.07.35.png
Mark Robinson (11:11:54) (in thread): > hmm .. and the repo you pointed to with the MATLAB code looks very different to the .M file in my snapshot.
Mark Robinson (11:13:44) (in thread): > you could also try writing to Xiaoyan (xiaoyan@cartana.se) .. she answered a few of my questions and is the author of the MATLAB code in the screenshot.
2021-04-21
Aishwarya Mandava (13:56:23): > @Aishwarya Mandava has joined the channel
2021-04-29
Jovan Tanevski (04:27:44): > @Jovan Tanevski has joined the channel
2021-05-11
Megha Lal (16:45:55): > @Megha Lal has joined the channel
2021-07-16
Lori Shepherd (12:42:46): > @Lori Shepherd has left the channel
2021-07-19
Leo Lahti (17:02:46): > @Leo Lahti has joined the channel
2021-08-26
Hervé Pagès (06:12:57): > @Hervé Pagès has left the channel
2021-09-06
Eddie (08:23:37): > @Eddie has joined the channel
2021-09-14
Jenny Drnevich (10:50:28): > Anyone have any ideas on how to rotate/mirror spatial images in either Seurat Objects or the new SpatialExperiment class?@Dario Righelli@Davide Risso@Lukas Weber@Helena L. Crowell. Both the image and the x,y coordinates will need to be modified. I know this problem has already been solved in R (https://www.rdocumentation.org/packages/spdep/versions/1.1-8/topics/Rotationorhttps://rdrr.io/github/ErasmusOIC/SMoLR/man/rotate_coord.htmlfor the x,y coordinates andhttps://rdrr.io/bioc/EBImage/man/spatial.htmlfor the image) but I have no idea how to apply these. I have 16 slices of mouse mid-brain and I need to get them all into the same orientation post-SpaceRanger. Any tips appreciated! - Attachment (rdrr.io): rotate_coord: Rotate Cartesian coordinates in ErasmusOIC/SMoLR: Single Molecule Localization Microscopy visualization and analysis in R > Rotate Cartesian Coordinates around a given center. Translate or flip cartesian coordinates. Stretch cartesian coordinates. - Attachment (rdrr.io): spatial: Spatial linear transformations in EBImage: Image processing and analysis toolbox for R > The following functions perform all spatial linear transforms: reflection, rotation, translation, resizing, and general affine transform.
2021-09-15
Aedin Culhane (05:45:50): > What cloud systems are institutes using to store/manage spatial/image data?
Sean Davis (10:02:41) (in thread): > Can you be more specific about what you are storing and what you want to do with it?
Lukas Weber (11:42:28) (in thread): > Hi@Jenny Drnevich, I have also had this issue, and have usually done it inside my plotting functions withscale_y_reverse()in ggplot. However maybe we need an alternative solution that modifies the SpatialExperiment itself. We will think about this - thanks for raising it.
Lukas Weber (11:42:41) (in thread): > Also tagging@Stephanie Hicks
Lukas Weber (11:50:30) (in thread): > I have opened a GitHub issue so we can think about this - thanks
2021-09-16
Henry Miller (18:35:17): > @Henry Miller has joined the channel
Henry Miller (18:35:24): > @Henry Miller has left the channel
2021-09-17
Vince Carey (08:03:23) (in thread): > Just guessing that doi: 10.1016/j.ejmp.2021.02.007 > would be relevant. It does not seem to be behind firewall. Includes a reference tohttps://link.springer.com/article/10.1007%2Fs11548-011-0625-xwhich is behind a paywall. However, there ishttps://en.wikipedia.org/wiki/Picture_archiving_and_communication_systemwhich may be more up-to-date. - Attachment (International Journal of Computer Assisted Radiology and Surgery): A PACS archive architecture supported on cloud… > International Journal of Computer Assisted Radiology and Surgery - Diagnostic imaging procedures have continuously increased over the last decade and this trend may continue in coming years,… - Attachment: Picture archiving and communication system > A picture archiving and communication system (PACS) is a medical imaging technology which provides economical storage and convenient access to images from multiple modalities (source machine types). Electronic images and reports are transmitted digitally via PACS; this eliminates the need to manually file, retrieve, or transport film jackets, the folders used to store and protect X-ray film. The universal format for PACS image storage and transfer is DICOM (Digital Imaging and Communications in Medicine). Non-image data, such as scanned documents, may be incorporated using consumer industry standard formats like PDF (Portable Document Format), once encapsulated in DICOM. A PACS consists of four major components: The imaging modalities such as X-ray plain film (PF), computed tomography (CT) and magnetic resonance imaging (MRI), a secured network for the transmission of patient information, workstations for interpreting and reviewing images, and archives for the storage and retrieval of images and reports. Combined with available and emerging web technology, PACS has the ability to deliver timely and efficient access to images, interpretations, and related data. PACS reduces the physical and time barriers associated with traditional film-based image retrieval, distribution, and display.
2021-09-25
Haichao Wang (07:20:18): > @Haichao Wang has joined the channel
2021-10-18
Qirong Lin (19:33:45): > @Qirong Lin has joined the channel
2021-11-08
Paula Nieto García (03:29:41): > @Paula Nieto García has joined the channel
2021-12-13
Sean Davis (18:26:21): > https://grants.nih.gov/grants/guide/notice-files/NOT-CA-22-013.html - Attachment (grants.nih.gov): NOT-CA-22-013: Notice of Special Interest (NOSI): Availability of Administrative Supplements to Harmonize / Map Existing Data to NCI Human Tumor Atlas Network Data Standards > NIH Funding Opportunities and Notices in the NIH Guide for Grants and Contracts: Notice of Special Interest (NOSI): Availability of Administrative Supplements to Harmonize / Map Existing Data to NCI Human Tumor Atlas Network Data Standards NOT-CA-22-013. NCI
2022-01-03
Kurt Showmaker (17:05:29): > @Kurt Showmaker has joined the channel
2022-01-13
Lambda Moses (22:44:14): > Is there something likesfforDataFrameascolDataofSpatialExperiment? I really likesfand how it acts like a tibble. This is relevant because the geometries can be relevant to spatial gene expression analyses, and I find traditional geospatial packages such assfandspdepimmensely helpful. For instance, thesfgeometries are useful for cell segmentation polygons in highly multiplexed smFISH data, and for representing and plotting Visium spots as their actual size relative to tissue. Also,sfallows you to set and convert units, and has fast operations like checking intersections (and different types of intersections) among geometries. Those operations are fast because of spatial indexing. Intersections are useful, such as to see how many nuclei each Visium spot contains or intersects, and which nuclei are in which manually annotated histological region.
2022-01-14
Vince Carey (13:06:56): > @Stephanie Hicks@Dario Righelli^^ … is this concept clear enough and if so do you see a path to doing some development in this direction? I have some familiarity with sf but am pretty booked up.
Dario Righelli (13:41:58): > Thanks@Vince Careyfor pointing this out. > I started to look into thesf(andsp) package(s) a while ago, but kinda lost myself in their huge documentation world, I think it could be a good thing to implement something for supporting these packages and facilitating the interaction with their data structures. > Maybe, if@Lambda Mosesagrees, we can schedule a meeting to better discuss which features could be helpful for the spatial omics world…
Lambda Moses (13:42:04): > I’m performing many analyses withsf, and I can collaborate
Dario Righelli (13:43:06): > That could be really helpful thanks!
Dario Righelli (13:44:07): > what about a meeting next week to discuss these analyses?
Kasper D. Hansen (14:14:20): > If we don’t already have support for this, I certainly think we should. But I would also say we need a different class for this, at least at first. With different segmentations you could imagine different matrix dimensions, and I think that will complicate things a lot
Ludwig Geistlinger (16:31:48): > Along the lines of different segmentations / different matrix dimensions: > > Where do we store alternative segmentations for one dataset at a time? > > ThealtExpsslot of aSummarizedExperiment-derivative is thought for alternative > experiments over a distinct set of features for the same samples. But here we have the > opposite situation, ie we have a distinct set of samples/cells for the same features. > > Is this just a separateSpatialExperimentwith potentially duplicating information > (incl. the images!) or a use case for a transposedaltExpsslot? > > I am asking because I have this situation with some recent MERFISH data, but from a > discussion with@Helena L. Crowelland@Dario Righelliearlier today that doesn’t seem to be > an uncommon situation, and is eg also encountered for different binnings in Stereo-seq > data.
Kasper D. Hansen (16:43:09): > I would advocate for a new class. I think from a data storage point, trying to unify these specific cases is going to be kind of pointless.
Lambda Moses (16:45:58): > I think probably yet another class for spatial point process analyses, for locations of cells and locations of transcript spots. The classic spatial point process packagespatstathas classes that are very different fromsf
2022-01-15
Helena L. Crowell (06:55:35): > Adding my (very random) thoughts to@Ludwig Geistlingerpoint regarding same-features-different-“cells”: > * I have tried two different approaches in my analyses: > > * storing separate binnings/segregations in a single object, i.e., one assay, tracking the bin/seg in thecolData(this is nice for plotting, but tricky for any downstream stuff since counts are essentially duplicated) > * having a list of objects, with duplicated shared metadata, e.g., images, or non-QC feature metadata (this is annoying but somehow less error-prone, I feel) > > * Now, I don’t like using the S-word, butSeuratdoes the latter I think (e.g., operations are carried out on a list - one object per sample - that is at some point joint into a single object > * From a devel perspective, putting all into one (e.g., via a slot analogues toaltExpsbut transposed (features must match, but not samples)), could be a possibility (in principle). However, if I think about analysing this, I’d eventually loop through thealtExpsinstead of the list I already have. So, I’m not sure this would make things “easier”, but it would certainly be “prettier” to have an object vs. a list.
2022-01-17
Shila Ghazanfar (04:18:55): > I think the problem of alternative segmentations could do with a new class - the challenge for a SpatialExperiment is that with different segmentations you have induced different “samples”, and so it’s unclear what the unit of observation is.. i’ve dealt with this for molecule-resolved data (seqFISH/MERFISH) by assigning a ‘assigned segmentation’ factor level to a molecules data frame, each row corresponding to an observed molecule with it’s own spatial coordinate and other factors (gene ID etc.)- then getting a counts matrix by summing over the segmentaiton and genes factors. I could imagine some function where you input a molecules dataframe and segmentation mask and output a SpatialExperiment with assays either simple matrix of counts or a bumpymatrix keeping the molecule level info
2022-01-27
Ludwig Geistlinger (10:26:18): > We would like to invite everyone interested in working with spatial transcriptomics > data to a seminar by Helena Crowell (@Helena L. Crowell) and Dario Righelli (@Dario Righelli) > on Mon, Jan 31, 10 AM EST / 4 PM CET. The seminar will provide an introduction to > current technologies and recent developments in Bioconductor for analyzing such > data. As such it is primarily intended for everyone who would like to get up to speed > with the current state of the art of obtaining and analyzing spatial transcriptomics > data in Bioconductor, but we of course also invite the spatial experts and veterans here > on the channel interested in adopting and building on top of these recent developments > to join the discussion. > > When: Mon, Jan 31, 10-11 AM EST / 4-5 PM CET > Where:https://harvard.zoom.us/j/97173440183?pwd=eHI1ODRub0p5NGNEZncwU0lURlJjdz09
Peter Hickey (19:35:29) (in thread): > Could this please be recorded and shared? 1am my time ain’t happening:sweat_smile:
Ludwig Geistlinger (19:57:25) (in thread): > Sure, we’ll record and share. Good point, thanks!
2022-01-28
Megha Lal (11:13:01): > @Megha Lal has left the channel
2022-02-07
Ludwig Geistlinger (10:57:32) (in thread): > Just a heads up that the slides and the recording of the spatial transcriptomics seminar are now up on the bioc website under course materials (https://www.bioconductor.org/help/course-materials/) and the recording is up on YouTube (https://www.youtube.com/watch?v=g-RWBHqKj6M). Thanks@Helena L. Crowelland@Dario Righellifor a fantastic overview, and thanks@Kayla Interdonatofor getting these materials up on the website / youtube!
Peter Hickey (16:19:54) (in thread): > Thanks Ludwig!
Ludwig Geistlinger (16:20:25) (in thread): > Sure thing!
2022-02-21
Peter Hickey (21:11:47): > What’s best practice for storing multiple Visium samples: 1SpatialExperimentper sample (capture area) or 1 per slide (e.g., 4 samples in the sameSpatialExperiment)? > I initially went with 1/slide but I noticed that the colnames of the resulting object (constructed withSpatialExperiment::read10xVisium()) are not unique (in contrast to reading in multiple captures withDropletUtils::read10xCounts()). Perhaps not a big deal but gave me pause
Lukas Weber (21:57:42): > This is a good point and I have noticed this recently too. I believe 10x Genomics uses the same set of 4992 barcode IDs for all capture areas. I disambiguated the column names by pasting together sample IDs and barcode IDs with an underscore, when this was causing problems for me (in a shiny app). Maybe we should think about formalizing this inSpatialExperiment::read10xVisium().@Helena L. Crowell@Dario Righelliany thoughts?
Peter Hickey (22:02:50): > that’s what I planned to do, too; FWIWDropletUtils::read10xCounts()does this for the user (“If col.names=TRUE and length(sample)==1, each column is named by the cell barcode. For multiple samples, the index of each sample in samples is concatenated to the cell barcode to form the column name. This avoids problems with multiple instances of the same cell barcodes in different samples.”)
Peter Hickey (22:04:49): > beyond that immediate issue, anything else to bear in mind in terms of using 1 vs. list of manySpatialExperimentobjects? I feel like I’d rather be working with 1 object where possible (this is largely what I do for SCE-based stuff)
Peter Hickey (22:05:57): > a colleague tells me Seurat go with a model of 1 sample = 1 object, which sounds cumbersome
Lukas Weber (22:09:20): > Yes, I prefer a single object containing all samples too, and have been working this way with our data. I think the main thing to check is that thesample_idcolumns incolDataandimgDatahave loaded correctly
Lukas Weber (22:16:45): > Also note that we have deprecated thespatialDataslot in the latest version 1.5.2 of the SpatialExperiment package, in response to feedback from users on GitHub as well as reviewer comments, i.e. that the use ofspatialDatawas ambiguous in terms of what to store there vs.colData. The columns that were previously inspatialDataare now all stored incolDatainstead, which makes things simpler (andspatialCoordsis still separate, as previously). The latest version (1.5.2) is in Bioc-devel or also works by installing from GitHub into Bioc-release. (I have switched to this version now to make things easier for my existing objects at the next Bioc-release.)
2022-02-22
Dario Righelli (03:22:28): > Mmm thanks@Peter Hickeyfor reporting this, my first idea to patch this is to extend the arguments ofread10xVisiumsimply by adding a...argument that will be the arguments for theread10xCountsthat we use for loading the counts. If theread10xCountsalready takes into account this issue…
2022-02-24
Mirana Ramialison (20:55:08): > @Mirana Ramialison has joined the channel
Amelia Dunstone (23:54:44): > @Amelia Dunstone has joined the channel
2022-03-08
Andrew McDavid (13:57:37): > @Andrew McDavid has joined the channel
2022-03-15
Jeanette Johnson (12:04:46): > @Jeanette Johnson has joined the channel
2022-03-21
Pedro Sanchez (05:02:55): > @Pedro Sanchez has joined the channel
2022-03-24
Ludwig Geistlinger (12:48:51): > We would like to invite everyone interested in working with spatial omics > data to a seminar with Nils Eling (@Nils Eling) on Mon, Mar 28, 10 AM EST / 4 PM CEST. > The seminar will focus on visualization and analysis of highly multiplexed imaging data, > and will provide an overview of current technologies and recent developments > in Bioconductor for analyzing such data. > > When: Mon, Mar 28, 10-11 AM EST / 4-5 PM CEST > Where:https://harvard.zoom.us/j/97173440183?pwd=eHI1ODRub0p5NGNEZncwU0lURlJjdz09
2022-03-28
Friederike Dündar (10:57:01) (in thread): > Thanks for sharing! > Can you add the details for next week’s seminar, too?
Ludwig Geistlinger (11:06:14) (in thread): > Sure! We will give a heads up later this week here. I can also add you to the mailing list if interested.
Friederike Dündar (11:07:01) (in thread): > Yes to both:slightly_smiling_face:
Ludwig Geistlinger (11:07:22) (in thread): > Alright, just DM me your email
Friederike Dündar (11:08:00) (in thread): > The one in my profile here will do
2022-03-31
Nicole Ortogero (22:27:36): > @Nicole Ortogero has joined the channel
2022-04-01
Ludwig Geistlinger (08:54:36) (in thread): > Just a heads up that the slides and the recording of the spatial image analysis seminar > are now up on the bioc website under course materials (https://www.bioconductor.org/help/course-materials/) > and the recording is up on Bioconductor’s youtube channel (https://www.youtube.com/watch?v=8pgbq5bQhxw). > Thanks@Nils Elingfor a very informative overview, and thanks@Kayla Interdonatofor getting these materials online. - Attachment (YouTube): Visualization Analysis IMC Data - March 28, 2022
Ludwig Geistlinger (08:56:47): > Also: for the next seminar we will have Giovanni Palla and Fabian Theis, who will be > presenting work onSquidpy: a scalable framework for spatial omics analysisWhen: Mon, Apr 04, 10-11 AM EST / 4-5 PM CEST > Where:https://harvard.zoom.us/j/97173440183?pwd=eHI1ODRub0p5NGNEZncwU0lURlJjdz09
Nils Eling (08:58:15) (in thread): > Thanks for organizing and uploading the material!
Ludwig Geistlinger (09:01:38) (in thread): > Thanks, Nils!
ImranF (09:06:35): > @ImranF has joined the channel
2022-04-21
Ynnez Bestari (17:43:35): > @Ynnez Bestari has joined the channel
Genoa Polumbo (17:45:46): > @Genoa Polumbo has joined the channel
2022-04-25
Indrik Wijaya (02:45:39): > @Indrik Wijaya has joined the channel
2022-04-27
Ynnez Bestari (10:26:27): > We would like to invite everyone interested in working with single-cell and spatial omics data to a seminar with Katy Börner on constructing a reference atlas for mapping healthy cells in the human body within the NIH Human BioMolecular Atlas Program. Please see below for meeting time and link: > Speaker: Katy Börner, Victor H. Yngve Distinguished Professor of Intelligent Systems Engineering, Indiana University > Title: Human Reference Atlas Construction and Usage > When: May 2nd, 2022 4:00-5:00 PM Eastern Time (United States and Canada) > > Join Zoom meetinghttps://harvard.zoom.us/j/97173440183?pwd=eHI1ODRub0p5NGNEZncwU0lURlJjdz09Password: 421790 > > Join by telephone (use any number to dial in)+1 929 436 2866+1 312 626 6799+1 301 715 8592+1 346 248 7799+1 669 900 6833+1 253 215 8782International numbers available:https://harvard.zoom.us/u/ausGPPR0POne tap mobile:+19294362866,,97173440183# US (New York) > > Join by SIP conference room system > Meeting ID: 971 7344 018397173440183@zoomcrc.com
2022-05-02
Ynnez Bestari (16:14:02): > Hello all, the seminar with Dr. Katy Börner is now live. Please join us here: Join Zoom meetinghttps://harvard.zoom.us/j/97173440183?pwd=eHI1ODRub0p5NGNEZncwU0lURlJjdz09Password: 421790
2022-05-03
Ray Su (06:55:53): > @Ray Su has joined the channel
2022-05-10
Genoa Polumbo (10:55:14): > Interested in working with single-cell and spatial omics data? Please fill out a brief survey as we construct HuBMAP: a reference atlas for mapping healthy cells in the human body within the NIH Human BioMolecular Atlas Program.https://app.smartsheet.com/b/form/0c3f6d63a3784b648a4083db9bbdcb56
2022-05-25
Isaac Virshup (07:53:02): > @Isaac Virshup has joined the channel
Sanket Verma (17:44:17): > @Sanket Verma has joined the channel
2022-05-31
Vince Carey (12:12:19): > Who<!here>is interested in zarr? It seems to me that the way to start on an R/bioc solution for working with zarr is to use basilisk/reticulate and build a client package for the zarr python module. After exploring use cases and benchmark results we can determine whether to go forward with direct interface to z5 or other approaches.
Marcel Ramos Pérez (12:13:43): > Here is a potential starting pointhttps://github.com/Bioconductor/ZarrExperiment
Vince Carey (14:51:32): > Yes, that predates basilisk though, so much of the installation-related code would be replaced.
Vince Carey (15:42:17): > Now I see in the issues thread there that basilisk was discussed for that package some 2.5 years ago. We’d really like to see some current use cases.
2022-06-01
Ludwig Geistlinger (12:23:27): > Hi@Vince Carey- adding some thoughts to this: there is a lot of interest in zarr in spatial / imaging world, and I guess it’s relevant for many folks here on the channel, even though most will be on the applied side (incl myself). Tagging@Dario Righelli@Helena L. Crowell@Nils Eling@Shila Ghazanfar@Lambda Mosesas I think they might be interested. > > On the other hand, some more technical folks have an eye on thishttps://github.com/zarr-developers/community/issues/18and it seems@Mark Kellerhas put some initial infrastructure in place:https://github.com/keller-mark/pizzarr. There seems to be also some support through the NetCDF library (ncdf4 package on CRAN). It would be good to know what others have to add especially with regard to using zarr over hdf5@Isaac Virshup@Luke Zappia@Peter Hickey@Hervé Pagès@Martin Morgan. > > With regard to current use cases, I have some Merfish images fromhttps://www.nature.com/articles/s41587-021-01044-wup on ExperimentHub (EH7544 and EH7545). Those are 9-channel z-stack tiffs, each 500 MB on disk, ie they still fit into memory on eg a mac book, but would maybe make for some good lightweight examples to play around with, currently they are served as EBImage objects. > > If we are looking for more challenging use cases, we could eg take the CyCIF tissue microarray TMA11.ome.tif (64 channels, 240 GB on disk) fromhttps://doi.org/10.1038/s41592-021-01308-y. This one is available from synapse (links can be found underhttps://mcmicro.org/datasets/). I have yet shied-away of making this one accessible via ExperimentHub as it seemed to go beyond dataset sizes envisioned for ExperimentHub (but can upload if prompted). > > As for wrapping around an existing python implementation: I might not be 100% qualified to comment, but just mentioning that we’ve lately run into some portability issues with the basilisk approach in a deployment setting (Shiny server / RStudioConnect). It seems that basilisk runs a conda env with dependencies into reticulate, yet deployment platforms such as RStudioConnect do not support apps or libraries that generate their own environments; According to folks from RStudio, RSC seem to require apps to give up all the environment variables to RSC not manage them on its own.@Tyrone Leecan comment on this one and just mentioning it here to make the case that a direct interface will likely be more robust/portable.
Mark Keller (12:23:43): > @Mark Keller has joined the channel
Luke Zappia (12:23:43): > @Luke Zappia has joined the channel
Hervé Pagès (12:23:43): > @Hervé Pagès has joined the channel
Tyrone Lee (12:23:43): > @Tyrone Lee has joined the channel
Hervé Pagès (12:49:51) (in thread): > What are the potential benefits of zarr over tiledb? Just curious as I don’t have any direct experience with either one.
Mark Keller (13:06:49) (in thread): > One reason that relates to my work withweb-based visualization toolsis that as far as I can tell there is no open source JS implementation for reading tileDB data (there is a JS library for thetileDB cloudpaid service). However there are multipleJS implementationsfor reading Zarr data
2022-06-02
Luke Zappia (02:38:47) (in thread): > I know there has been interest around using zarr for single-cell data in the Python scverse world and it has been used for some things but I think there are limitations that stop it from taking over from HDF5 at the moment.@Isaac Virshupcan probably explain more.
Nils Eling (03:16:37) (in thread): > I’m not very familar with using zarr but if OME-NGFF becomes a new standard for storing bioimaging data it would definitely be nice to have zarr reader functions in R.
Shila Ghazanfar (04:34:08) (in thread): > +1 for OME-NGFF, I mentioned this discussion to Constantin Pape who’s passed along to Josh Moore. Incorporating R readers for these file formats would be very useful indeed. Currently, i view image data in Fiji/Napari and then abstracted spatial omics data in R/Shiny, and it would be great to enable joint visualisation & analysis
Josh Moore (06:03:59): > @Josh Moore has joined the channel
Josh Moore (06:16:51) (in thread): > And to close the circle, hi:wave:. To the tiledb question, I tend to think of the trade-off as simplicity vs. power, but ultimately I think we want to work toward interoperability of the two. e.g. from talking to Stavros, a common operation might be to pull an at rest Zarr dataset into tiledb in order to do more advanced operations. (That might be less true for image data, but happy to hear opinions.)
Josh Moore (06:19:46) (in thread): > OME & IDR will continue to move images into NGFF (https://idr.github.io/ome-ngff-samples/) simply because we need a way and a place to keep up with growth. I imagine having direct access in R will be ideal, but of course, if we can help anyone with a conversion workflow (e.g. getting the images into OME-TIFF) let us know. To the extent that the dataisin Zarr and you also have ann- or mudata’s in Zarr, then personally I’m really excited about the idea of holistically linking those datasets.</over>
Isaac Virshup (09:02:03) (in thread): > @Luke Zappia, R not having a good way to read Zarr is the biggest reason I would currently recommend HDF5 for saving AnnDatas . Other than that, to me, Zarr is a more simplified, modern HDF5. > > I’d second Josh’s comments on zarr vs tiledb. To me, zarr is much more simple, which makes high performance low level access much easier.
Tyrone Lee (12:09:27) (in thread): > There is a suite of tools to integrate and execute python code within R and vice versa.Reticulatehas a generic method for converting common python objects into R objects. In my experience, I can convert most anndata or zarr objects into numpy arrays then into R. There is some cleanup needed with metadata but its serviceable.
Tyrone Lee (12:17:47) (in thread): > At CCB we did have issues with basilisk which seems to be the de-facto means for standardizing python requirements in R packages. This library creates a conda env to run reticulate, which doesn’t play well with some platforms likeR Studio Connect. I don’t know if anyone knows but is there a reason why it seems that bioconductor libraries that depend on basilisk ( or just zellkonverter and vitessce) all use python3.7.7 ?
Mark Keller (12:41:05) (in thread): > It looks here like if python version is not specified in the basilisk conda environment definition, then it will use the version coming from the system python binaryhttps://github.com/LTLA/basilisk/blob/9c543869dfaba3f57d0db1a9431af326fe4bdc32/R/setupBasiliskEnv.R#L85
Mark Keller (12:43:18) (in thread): > https://github.com/LTLA/basilisk.utils/blob/master/R/getCondaBinary.R#L28
Mark Keller (12:46:01) (in thread): > Or it could be coming from the usage of this particular Miniconda installerpy37_4.8.3https://github.com/LTLA/basilisk.utils/blob/master/R/installConda.R#L97
Tyrone Lee (12:51:55) (in thread): > I had a feeling it was because of the miniconda installer; that developers were using whatever version was given by the library. I wasn’t sure if there was any collective agreement to use this particular python version for compatibility or something.
Vince Carey (15:45:13) (in thread): > I know of no agreement on python version – I changed the setting on L97 of basilisk.utils/R/installConda.R noted by Mark above to py39_4.12.0 and all was well for the BiocSklearn package with > > added / updated specs: > - h5py=3.6.0 > - pandas=1.2.4 > - python=3.9.12 > - scikit-learn=1.0.2 > > announced during the installation
Vince Carey (15:48:08) (in thread): > Given that problems with R Studio Connect are observed with basilisk, I think it is possible to consider basilisk discipline as “opt in” – a package can be designed so that python usage is, by default, managed by reticulate, with the attendant risks, but when stronger guarantees are needed, basilisk rules the show. I don’t think there’s a clear example or pattern to do this – it can be accomplished in various ways. But at this point we should not regard basilisk as an obstacle.
2022-06-03
Luke Zappia (02:39:08) (in thread): > Some details on how****{zellkonverter}****handles this. I don’t think there is any reason you have to use a specific Python version. The default ****{basilisk}**** environment in the latest release using Python 3.8 (we also allow you to select an older environment which uses Python 3.7 andanndatav0.7.6). We have keep the conversion functions (SingleCellExperiment<->AnnData) separate from the functions for reading/writing H5AD files with only the H5AD function wrapped in a****{basilisk} ****call. This means that if you usereadH5AD()/writeH5AD()the code is run in a****{basilisk}****environment but if you are familiar enough with****{reticulate} ****to get a pointer to a PythonAnnDatayou can useAnnData2SCE()/SCE2AnnData()directly in whatever your current****{reticulate}****environment is (hopefully that makes sense). This gives the option of having everything managed for you if you don’t want to mess around with environments or using your own if you are comfortable doing that. > > We have also just submitted a CZI grant with people from MuData and Bioconductor core around building a robust R interface for H5AD files so you don’t need to mess around with Python at all. If that is successful I can see potential for extensions to zarr based formats later on but we will have to wait and see.
Tyrone Lee (12:17:36) (in thread): > I agree with@Vince Careythat basilisk should not be an obstacle for data portability. The exception is within the RStudioConnect platform due to its strict enforcement of python and R versions and virtual environment control. We had trouble merely publishing apps to the platform if the basilisk python ver. does not match the RSC python ver. We would like our apps for RStudioConnect to have more functionality; they could execute functions that can read/write to H5AD/anndata/zarr. From our discussions with RStudio engineers, this is only possible if there was an R interface for these data types. Currently right now we’re using scripts outside of these apps to manipulate data and only using the RSC apps to visualize. > I understand this is a very particular case here and probably has little relevance so my apologies for derailing discussion. I just brought it up for curiosity’s sake and to get a better explanation for the RSC solution engineers puzzling over our issues with the platform.
2022-06-21
Stephanie Hicks (21:08:17): > Congratulations@Shila Ghazanfaron your CZI Data Insights Award!:tada:https://chanzuckerberg.com/science/programs-resources/single-cell-biology/data-insights/multiscale-data-integration-for-single-cell-spatial-genomics/ - Attachment (Chan Zuckerberg Initiative): Multiscale Data Integration for Single-Cell Spatial Genomics - Chan Zuckerberg Initiative
Peter Hickey (21:34:38): > Also to@Federico Mariniand@Katharina Imkellerforhttps://chanzuckerberg.com/science/programs-resources/single-cell-biology/data-insights/unraveling-immunogenomic-diversity-in-single-cell-data/:tada: - Attachment (Chan Zuckerberg Initiative): Unraveling Immunogenomic Diversity in Single-Cell Data - Chan Zuckerberg Initiative
Stephanie Hicks (22:14:16): > OH NICE! Yes, absolutely CONGRATULATIONS@Katharina Imkeller@Federico Marini! ha, I’ve been reading through all the awardees tonight and hadn’t gotten that far down the list yet!:tada:
2022-06-22
Federico Marini (08:39:29): > @Federico Marini has joined the channel
Federico Marini (08:41:44): > Thanks everyone - Quite excited to get the computational-hands dirty on this:slightly_smiling_face:Credits to the original idea of@Katharina Imkeller, and hooray for productive meetings that are indeed a strong component of in-person meetings:tada:
Federico Marini (09:24:57): > … and there’s also@Tim Trichein the pool of awardees:slightly_smiling_face:
Shila Ghazanfar (20:51:23): > Thank you all and MASSIVE CONGRATS to@Federico Marini@Katharina Imkeller@Tim Triche!!! It’s so humbling to have this supportive community and I’m looking forward to bringing more on board for understanding molecule-resolved spatial omics with Bioconductor!!
2022-06-23
Jenny Drnevich (10:04:18): > Anyone have experience with multiple tissue sections on one Visium capture area? I recently had one where they embedded 9 mouse dorsal root ganglia in the same block, then sliced and hybridized. Some of those turned out too small (~20 spots total) to do much with the data so I didn’t get too far into the analysis. We now have someone else with 1 mm tissue biopsy punches and they are hoping to embed 3-4 per block/capture area on a full slide. It seems like it wouldn’t be that hard to pull the spot barcodes for each tissue section in order to assign individual sample IDs, and many of the processing/analyses steps shouldn’t be any different than processing more than one capture area together. I do not think they are going to want any sophisticated spatial analyses, just clustering and then DE testing between clusters and also between sets of samples. Am I missing any major considerations that would make analyzing multiple tissues per capture a very bad idea? Seems like this could become much more common once the Visium HD comes out…
Stephanie Hicks (13:54:25) (in thread): > Tagging@Lukas Weberwhoisworking withsomesimilardataincasehehasthoughts.
Lukas Weber (13:57:33) (in thread): > Hi, yes we have a dataset like this right now too, with up to 2-3 parts per Visium capture area. The parts are from sequential slices in a tissue block, so the captured information is very similar. I have recorded the part IDs in our SpatialExperiment objects, but I haven’t used it in most of the analyses, i.e. I am treating the whole Visium capture area as a single sample during the analyses. I think this makes sense in our case, so the multiple parts are simply a way to capture more information from the same samples
Jenny Drnevich (15:26:41) (in thread): > You’re right, treating the technical replicates as all the same sample seems like the easiest thing to do. I’ll have to ask if they are doing multiple punch biopsies per tumor or only 1 per tumor…
Lukas Weber (18:29:42) (in thread): > Yes, if the pieces are from separate biological samples, then I think this makes things more difficult, and in that case it may be better to treat the parts as samples. Some exploratory visualizations could be helpful to see where the main variation lies, e.g. a PCA plot colored by sample vs. colored by part
2022-06-24
Jenny Drnevich (10:05:54) (in thread): > BTW - I was just sent your recent Giga Science paper on genetic demultiplexing of pooled single cell samples. We are interested in trying it here to save money. Probably not cancer samples, but I assume the same concept will work as long as the samples are genetically enough distinct? So not replicates from the same cell line…
2022-07-01
Ludwig Geistlinger (10:46:40) (in thread): > As a follow-up to this thread, we have now CyCIF image data in zarr format up on Bioconductor’s OpenStorageNetwork for everyone interested in experimenting more with zarr format in R. Please join the#zarrchannel for details and if interested in joining the effort.
2022-07-04
Andrew J. Rech (19:45:20): > @Andrew J. Rech has joined the channel
2022-07-05
Tyrone Lee (16:55:19): > Bit of an open question but I think@Josh Moorementioned in a thread. I’m trying to get MERFISH images fromhttps://datadryad.org/stash/dataset/doi:10.5061/dryad.jm63xsjb2which are two single-channel tiffs with 9 z-layers, one dapi stain and another for Na+/K+-ATPase. I was curious if there was a way to parse the cell-label and cell boundary data with images into OME-tiff or OME-ngff so I can visualize the molecule data on top of the tiff images. > These aren’t 10x Visium slides so I’m at a loss at how to proceed from here.
Shila Ghazanfar (21:12:19) (in thread): > Hi Tyrone! what format is the cell-label and boundary data? tiff? i was able to convert seqFISH tiff images to OME-NGFF and can now visualise the molecule image data and segmentation in napari using the napari-ome-zarr plugin (all with tremendous input from Constantin Pape:pray:- see screenshot). it required python scripting to create the OME-NGFF files, happy to chat more and share > > somewhat aside, is it that this data might already be on the IDR and in such a format? maybe@Ludwig Geistlingeris familiar? cheers, - File (PNG): image.png
2022-07-06
Nils Eling (05:15:45) (in thread): > Not sure if this helps: we have a notebook that reads in and formats multichannel images and segmentations masks (both TIFF format) for composite visualisation and cell outlining in napari:https://github.com/BodenmillerGroup/IMCDataAnalysis/blob/biorxiv_submission/code/mask_overlay/mask_overlay.ipynb
Ludwig Geistlinger (11:53:59) (in thread): > Thanks Shila and Nils! Shila, Tyrone is actually working with me on the Merfish data. Can you point us to any tools / code that you used to convert your seqFISH tiff images to OME-NGFF?
Ludwig Geistlinger (11:56:44) (in thread): > As Tyrone pointed out our image data are two single-channel tiffs with 9 z-layers, one dapi stain and another for Na+/K+-ATPase. All other information (cell labels, coordinates, polygons, …) comes from plain text files and we have it wrapped in aSpatialExperiment. What Tyrone aims to do is taking these plain text annotations and code them into an OME header for the plain tiffs that we are having in order to visualize things with Napari or Vitessce/Viv.
Tyrone Lee (13:03:42) (in thread): > To quote the author of the data package: > “MERFISH was done against cryosections of the mouse ileum using standard protocols. Imaging was performed on a purpose-built microscope and flow system (not 10X vivisium nor any proprietary microscope system). Identification of RNAs was performed with an open-source analysis package (github.com/ZhuangLab/MERFISH_analysis). Cell boundaries were done with Baysor analysis and cellpose extracted into flat csv files.” > > Basically just .tiffs and .csv files with all the molecules, cell types, centroids, segmentation, and cell-boundaries. To me the simplest way would be to parse into an XML header but there’s no ready made tools available it seems (bioformat doesn’t apply here). Mark Keller does have a wrapper that goes from SPE -> anndata -> zarr but it seems to not handle images that are not Vivisim slides.
2022-07-07
Clara Pereira (14:27:21): > @Clara Pereira has joined the channel
2022-07-15
Ashley Robbins (15:23:48): > @Ashley Robbins has joined the channel
2022-07-24
Farhan Ameen (20:43:33): > @Farhan Ameen has joined the channel
2022-07-28
Will t (09:46:32): > @Will t has joined the channel
2022-08-04
László Kupcsik (21:59:13): > @László Kupcsik has joined the channel
2022-08-08
Natalie charitakis (20:55:45): > @Natalie charitakis has joined the channel
Denis Bienroth (22:05:03): > @Denis Bienroth has joined the channel
Hieu Nim (23:33:11): > @Hieu Nim has joined the channel
2022-08-11
Rene Welch (17:16:22): > @Rene Welch has joined the channel
2022-08-15
Michael Kaufman (13:13:34): > @Michael Kaufman has joined the channel
2022-08-30
Lisa Breckels (07:20:38): > @Lisa Breckels has joined the channel
2022-09-04
Gurpreet Kaur (15:00:19): > @Gurpreet Kaur has joined the channel
2022-09-13
Alex Schrader (17:17:14): > @Alex Schrader has joined the channel
2022-09-14
Robert Ivánek (05:19:31): > @Robert Ivánek has joined the channel
2022-09-23
Iivari (08:40:53): > @Iivari has joined the channel
2022-09-27
Jennifer Holmes (16:15:10): > @Jennifer Holmes has joined the channel
2022-10-06
Michael Love (12:24:32): > @Michael Love has joined the channel
Michael Love (12:25:36): > @Lukas Weberfor OSTA book, is there a place where you go into the questions that spatial helps to answer? there are so many great resources: datasets, preprocessing tools and workflows. But if a student is brand new, is there a place where Bioc lays out the scientific questions?
Lukas Weber (12:26:00): > Thank you, I don’t think we have a specific channel right now, but I can make one!
Michael Love (12:26:32): > oh and sorry i meant, place = section in the OSTA book
Michael Love (12:27:35): > like, the data is more rich and honestly interesting than gene expression matrices. but even with simple gene expression, we can say, ok so we are going to learn about gene regulation by making a perturbation and then seeing what genes apparently change in their abundance
Michael Love (12:28:54): > the context is, i know a lot of students are really interested in getting into spatial txomics, and i think they would benefit from hearing from you (the experts) how the methods and data answers particular questions
Shila Ghazanfar (12:29:13): > hi mike, i think there is some discussion of such in this latest review:https://www.nature.com/articles/s41587-022-01448-2Figure 4 but it’s not primary focus of the paper - Attachment (Nature): The expanding vistas of spatial transcriptomics > Nature Biotechnology - Spatial transcriptomics workflows, metrics and limitations are reviewed and discussed.
Michael Love (12:29:43): > maybe this or other papers like it could get linked out from the OSTA in the intro somewhere?
Lukas Weber (12:30:04): > Oh, yes that is a good question. Right now in OSTA it is structured more along the lines of explaining analysis tools for someone who already knows about the questions that they are planning to address. So yes, maybe some more of the scientific context could be helpful
Lukas Weber (12:30:54): > Thanks@Shila Ghazanfar@Michael Love, yes maybe one way to do this would be to include some more links to review papers like this one. There have been a couple of nice review papers recently, which provide a good introduction / overview
Stephanie Hicks (12:31:34) (in thread): > great idea@Michael Love! We should definitely do this@Lukas Weber:slightly_smiling_face:
Lukas Weber (12:32:19): > Maybe an intro chapter on “scientific context” / “scientific background” or similar, with some quick discussion and links to these types of review papers
Michael Love (12:33:43): > i think this would help make Bioc (even more so) the main place for comp bio / biostat students to learn about the whole field
Michael Love (12:34:14): > like, a dozen years ago, i learned about gene expression and R/Bioc at the same time by reading the Bioc books
Shila Ghazanfar (12:35:50): > yes i think thats a great point mike, perhaps its time for a ‘spatial omics’ version ofhttps://genomebiology.biomedcentral.com/articles/10.1186/s13059-020-1926-6 - Attachment (BioMed Central): Eleven grand challenges in single-cell data science - Genome Biology > The recent boom in microfluidics and combinatorial indexing strategies, combined with low sequencing costs, has empowered single-cell sequencing technology. Thousands—or even millions—of cells analyzed in a single experiment amount to a data revolution in single-cell biology and pose unique data science problems. Here, we outline eleven challenges that will be central to bringing this emerging field of single-cell data science forward. For each challenge, we highlight motivating research questions, review prior work, and formulate open problems. This compendium is for established researchers, newcomers, and students alike, highlighting interesting and rewarding problems for the coming years.
Michael Love (12:37:50): > Yes you have great influence when writing these resources, to shape how students think about things, beyond the data structure too@Shila Ghazanfarthis paper you linked is just what i was looking for “Broadly speaking, we see ST as best suited to answering three kinds of biological questions…”
Shila Ghazanfar (12:40:22): > maybe we can start to put some words and ideas together? in the meantime we can look for additional resources/papers (and add to OSTA intro)
Pedro Sanchez (12:45:30): > This sounds super interesting!!! In case it helps with more good ST papers,@Ramon Massoni-Badosawrote an amazing Twitter thread two days ago:https://twitter.com/rmassonix/status/1577238852397596673?s=20&t=xTH0gKfvXouHTVrIefMYWg - Attachment (twitter): Attachment > 3 key reviews to get you started with spatial transcriptomics:
Ramon Massoni-Badosa (12:45:36): > @Ramon Massoni-Badosa has joined the channel
Lukas Weber (13:44:17) (in thread): > Thank you! This is a useful list
2022-10-31
Chenyue Lu (10:05:45): > @Chenyue Lu has joined the channel
2022-11-06
Sherine Khalafalla Saber (11:21:37): > @Sherine Khalafalla Saber has joined the channel
2022-11-17
Shila Ghazanfar (18:34:08): > hi all - sharing across Isaac’s message from the zarr channel in case of interest. OME-NGFF relevant for imaging-based spatial omics datasets - Attachment: Attachment > Hey all, there’s a conversation on the image.sc forum about reading OME-NGFF zarr stores (espeically tables) from R. Would appreciate a perspective from bioconductor here! > > https://forum.image.sc/t/accessing-ome-ngff-esp-tables-from-r/74051
2022-11-18
Vince Carey (09:04:01): > @Ludwig Geistlingerwould you comment here? I am hoping to get more involved with this now that the release of 3.16 is complete, but it will be another couple of weeks til I can focus. We should have a road map. There is a bit of discussion of R-python interfacing that does not evidence awareness of basilisk. I think it helps avoid some of the version difficulties alluded to in the discussion. If there are C or C++ libraries deeper down that would be usable we should look at them.
Kasper D. Hansen (10:30:00): > I also read the discussion and I also think a basilik version would solve the issues.
Kasper D. Hansen (10:31:17): > I did not love thinking about a - supposedly high performance (?) - file format without a high quality C(++) library for parsing it. Or at least that was my impression from reading the issues. If this is a complicated file format and all they have is python stuff, I am wondering if this is really a good format.
Kasper D. Hansen (10:31:43): > I mean a good general-purpose format.
Kasper D. Hansen (10:32:42): > But leaving aside these editorial comments, the immediate path forward seems like one should use basilisk
Mark Keller (10:36:21) (in thread): > Zarr does have a C++ libraryhttps://github.com/constantinpape/z5There is further discussion here about ithttps://github.com/zarr-developers/community/issues/18#issuecomment-552783210 - Attachment: Comment on #18 Usage from R
Kasper D. Hansen (10:41:55) (in thread): > Super interesting. I noted the following comment on the webpage > > This library implements the zarr and n5 data specification in C++ and Python. Use it, if you need access to these formats from these languages. Zarr / n5 have native implementations in Python / Java. If you only need access in the respective native language, it is recommended to use these implementations, which are more thoroughly tested.
Kasper D. Hansen (10:42:28) (in thread): > To me, this suggests that the native python implementation is the one people use, whereas this is a “secondary” parser so to say.
Kasper D. Hansen (10:42:55) (in thread): > So this raises the question of how good this C++ library is, how well supported and how much we would want to build upon this.
Kasper D. Hansen (10:44:22) (in thread): > Do you know the relationship between the developers of this library and the main zarr python parser?
Mark Keller (10:51:23) (in thread): > I do not know the full history of the Zarr project but you can find all of the implementations herehttps://github.com/zarr-developers/zarr_implementations(the python implementation is herehttps://github.com/zarr-developers/zarr-python). The specification is also relatively simplehttps://github.com/zarr-developers/zarr-specs
Mark Keller (10:51:34) (in thread): > The governance structure is herehttps://github.com/zarr-developers/governance/blob/main/GOVERNANCE.md#zarr-steering-council
Vince Carey (10:56:13): > I’ll just note that there is a dedicated#zarrchannel in this slack, and I have pinned a discussion between bioc and zarr devs at CZI athttps://community-bioc.slack.com/archives/C03MKFSS7V2/p1663623032854839 - Attachment: Attachment > Here are notes on the CZI discussion: https://docs.google.com/document/d/1KpMbZbn789FPCMvsgKdIrHbSm74K_XuVY2XqCbzyJHs/edit?usp=sharing
Vince Carey (11:06:44): > Additionally, we havehttps://mghp.osn.xsede.org/bir190004-bucket01/index.html#TMA11/which includes several representations of a tissue-microarray for benchmarking.
Ludwig Geistlinger (12:27:57): > The question is a bit confusing. It starts being 1) about reading OME-NGFF into R and then it ends up being a question about 2) reading zarr archives into R. For 1) ie reading OME formats, a good starting point will be theRBioFormatspackage. For 2) using arecticulateapproach around python’szarrpackage as eg implemented inZarrExperiment, or one of the several (still rather experimental) approaches that Vince had collected in the document above would be my recommendation.
2022-11-21
Josh Moore (05:42:11) (in thread): > With the additionalhttps://github.com/ome/zarrreaderjar, RBioFormats should be able to read OME-Zarr (i.e. OME-NGFFs) now.
Ludwig Geistlinger (10:41:28) (in thread): > Interesting. Thanks for the pointer! I poked Andrezej about it:https://github.com/aoles/RBioFormats/issues/34 - Attachment: #34 Reading OME-Zarr (OME-NGFF) > A discussion in the Bioconductor slack channel was wondering whether RBioFormats could already be used to read OME-Zarr format. It seems that with the additional https://github.com/ome/zarrreader jar, RBioFormats would be able to read OME-Zarr (i.e. OME-NGFFs) now?
2022-11-22
Isaac Virshup (10:16:53) (in thread): > The original implementation was in python. There are now C implementations in GDAL and netcdf as well as a number of C++ implementations including tensorstore. > > If you’re interested in collaboration on a header only C or C++ library, I could put you in touch with other interested parties from GraphBLAS and UniData
2022-12-12
Umran (17:58:23): > @Umran has joined the channel
2022-12-13
Ana Cristina Guerra de Souza (09:01:32): > @Ana Cristina Guerra de Souza has joined the channel
Xiangnan Xu (18:33:05): > @Xiangnan Xu has joined the channel
2022-12-14
Lijia Yu (19:38:16): > @Lijia Yu has joined the channel
2022-12-20
Jennifer Foltz (10:41:41): > @Jennifer Foltz has joined the channel
2022-12-23
Pierre Gestraud (08:35:55): > @Pierre Gestraud has joined the channel
2023-01-09
Giovanni Palla (06:21:40): > @Giovanni Palla has joined the channel
Luca Marconato (11:30:45): > @Luca Marconato has joined the channel
2023-01-10
Tim Treis (05:44:24): > @Tim Treis has joined the channel
Vince Carey (10:46:24): > Can I suggest we drop/archive the channel spatialexperiment which has gone stale since last august?
Stephanie Hicks (11:30:12) (in thread): > That sounds good to me.@Helena L. Crowell@Lukas Weber– thoughts? i see@Dario Righellihas given a:thumbsup:
Lukas Weber (11:31:33) (in thread): > Hey, yes I think this would be ok. We can have any additional discussion in this channel instead if/when we need it
Lukas Weber (11:31:48) (in thread): > Also the other related one spatialexperiment-devel could be archived too
2023-01-20
Hervé Pagès (02:40:12): > @Hervé Pagès has left the channel
2023-01-21
Hien (16:04:33): > @Hien has joined the channel
2023-02-14
Lukas Weber (11:56:17): > Quick update regarding SpatialExperiment – our fix for the error in the Windows builds (discussed in#spatialexperiment-devel) has now passed and the builds and checks are green again. > > Also@Vince Careycould you please archive both the#spatialexperimentand#spatialexperiment-develchannels for us (as discussed above – since we haven’t been using those channels much lately)? (It looks like only workspace admins can archive channels, so the option in settings is grayed out for me.) We can continue/consolidate the discussion on future updates and extensions to SpatialExperiment in this channel instead. Thank you!
Sean Davis (12:24:57) (in thread): > Just archived. Thanks for the consolidation,@Lukas Weber!
Lukas Weber (12:25:14) (in thread): > Thank you!!
2023-02-22
michaelkleymn (01:44:34): > @michaelkleymn has joined the channel
2023-03-01
jeremymchacón (12:14:39): > @jeremymchacón has joined the channel
2023-03-02
Laurent Gatto (12:52:11) (in thread): > Just FYI - there’s arawrrpackage in Bioc. There might be some confusion with a/therarrpackage:slightly_smiling_face: - Attachment (Bioconductor): rawrr
Leonardo Collado Torres (13:59:40): > FYISpatialExperimentauthors. We just got some data from 10x using their internal spaceranger version that seems to include changes vs the publicly released version 2.0.0. - File (PNG): Screenshot 2023-03-02 at 12.58.52 PM.png
Leonardo Collado Torres (14:02:36): > annoyingly, they changed again the file name for the positions. In earlier versions it wastissue_positions_list.csv(plural, no header) then in 2.0.0 it becametissue_positions.csv(plural, with a header) and now in this newer version it’stissue_position.csv(singular, with a header). This will likely affecthttps://github.com/drighelli/SpatialExperiment/blob/7b8289fec1b94c37f91546e724bffc2be554bc10/R/read10xVisium.R#L130and well, causes problems when users manually re-name files (like a collaborator renamed them totissue_positions_list.csvand ran into an error due tohttps://github.com/drighelli/SpatialExperiment/blob/7b8289fec1b94c37f91546e724bffc2be554bc10/R/read10xVisium.R#L196) > > Error in `spatialCoords<-`(`**tmp**`, value = as.matrix(colData(spe)[i])) : > is.numeric(value) is not TRUE >
Leonardo Collado Torres (14:03:37): > we don’t know whether this change will make it into the next public release, but I thought you might want to have a heads up@Dario Righelli@Helena L. Crowell@Lukas Weber
Lukas Weber (14:29:59) (in thread): > Thanks@Leonardo Collado Torres, will look into this
Vince Carey (23:13:34): > Do we have any human connections to 10x software devs? It would be good to have some access to road map/planned changes instead of learning about them after the fact/through error. I could try to find a connection if there is none at this time.
Stephanie Hicks (23:16:28): > I know a few of them, but even so they have never been able to tell me in advance of these types of changes unfortunately.:confused:
2023-03-05
Shila Ghazanfar (05:47:05): > i have a few connections in australia, who have been able to reach out, so not particularly direct i’m afraid..
2023-03-06
Anthony Ramnauth (10:41:30): > @Anthony Ramnauth has joined the channel
2023-03-07
Monalika Patnaik (15:08:36): > @Monalika Patnaik has joined the channel
2023-03-08
Ibizugbe Merit (00:17:54): > @Ibizugbe Merit has joined the channel
Unyime William (01:15:44): > @Unyime William has joined the channel
Namusisi Sharon (09:57:15): > @Namusisi Sharon has joined the channel
2023-03-24
Boyi Guo (12:36:29): > @Boyi Guo has joined the channel
2023-03-26
Ellis Patrick (21:53:52): > Has anyone had any issues reading qptiff images with Rbioformats? Our images which are ~8GB on disk blow up to 180GB in memory, with ~800GB being used as the image is read. Any tips or feedback or general philosophies would be very much appreciated from@Nick R.and I before we sink more time into this.
Nick R. (21:53:55): > @Nick R. has joined the channel
2023-03-28
Phil Cheng (15:06:34): > @Phil Cheng has joined the channel
2023-03-29
Himel Mallick (16:56:57): > @Himel Mallick has joined the channel
2023-03-31
Ilaria Billato (08:58:59): > @Ilaria Billato has joined the channel
2023-04-06
Jenny Drnevich (10:13:36): > Posted on behalf of a friend on the bioinfo-core: > > Are you planning to attend ISMB in Lyon this year? The bioinfo-core COSI is planning a 20-30 minute panel about spatial transcriptomic data as part of our bioinfo-core workshop duringISMB/ECCB Lyon. > > If you are in a bioinformatics core facility or in a similar analytics role and struggling with an abundance of spatial transcriptomic data? Or maybe you manage or lead a core and have some experience and knowledge to share with the community? Or know someone who does? > Panel topics may include: > * current challenges with spatial data > * platform pros and cons or experiences > * best practices or things that worked > * tools for data analysis, integration, visualization > * strategies for finding time/resources to spend on development of new approaches > Please contact Radhika Khetani <rkhetani@hsph.harvard.edu> or Madelaine Gogol <MCM@stowers.org> if you are interested in participating in this panel. The organizers are planning to provide complimentary conference registration for our speakers and panelists.
2023-04-07
Vince Carey (06:29:11): > @Shila Ghazanfar@Ellis Patrick^^ will your groups be represented at ISMB and is this a relevant opportunity? …@Ludwig Geistlinger
Ellis Patrick (08:21:46) (in thread): > I was considering dropping in on my way to Bioc. Ive budgeted for ISMB, will submit an abstract, but haven’t cleared it with my wife yet. Would love to be involved though.
Ludwig Geistlinger (21:17:48) (in thread): > Yes it’s interesting and we are there anyhow. I’ll speak with Davide.
2023-04-10
Shila Ghazanfar (18:27:28) (in thread): > thanks for this vince! it wasn’t on my radar for this year but it would be great to contribute somehow
2023-05-17
Hassan Kehinde Ajulo (12:11:36): > @Hassan Kehinde Ajulo has joined the channel
2023-05-18
Davide Risso (01:35:52): > https://twitter.com/10xgenomics/status/1658936333929422850?s=46&t=dLb5okm7fr15BtVm3q7rGQ - Attachment (Twitter): 10x Genomics on Twitter > This morning, NanoString received a court-ordered injunction to stop selling CosMx instruments and reagents for RNA detection in Germany. We’re dedicated to ensuring your research continues. Learn more: https://t.co/2gZDvE9d82 (1/5)
Alan O’C (03:42:11): > Love that they frame it so positively, “we’re dedicated to ensuring your research continues” lol
Davide Risso (03:44:07) (in thread): > it sounds better than “we’re dedicated to making money and ensuring all competitors go under” doesn’t it?
Alan O’C (03:45:23) (in thread): > I am dedicated to making sure you get home on time, as I slash the tires on your car
Davide Risso (03:46:12) (in thread): > (on my expensive taxi)
Oluwafemi Oyedele (05:54:14): > @Oluwafemi Oyedele has joined the channel
2023-06-07
Laura Luebbert (13:35:13): > @Laura Luebbert has joined the channel
2023-06-09
Ahmad Alkhan (05:57:36): > @Ahmad Alkhan has joined the channel
Aedin Culhane (17:26:22): > @Ahmad Alkhan
2023-06-19
Ida Holopainen (11:21:18): > @Ida Holopainen has joined the channel
Meeta Mistry (20:27:24): > @Meeta Mistry has joined the channel
2023-06-25
Shila Ghazanfar (23:36:51): > hi spatial omics folks! we’re looking to improve upon ourreadMerscope()functionality in theMoleculeExperimentpackage - and are looking for examples of the data/folder structures coming off the MERSCOPE platform — if this is data you’re working with, please reply in the thread below or DM me:pray:thanks! - Attachment (Bioconductor): MoleculeExperiment > MoleculeExperiment contains functions to create and work with objects from the new MoleculeExperiment class. We introduce this class for analysing molecule-based spatial transcriptomics data (e.g., Xenium by 10X, Cosmx SMI by Nanostring, and Merscope by Vizgen). This allows researchers to analyse spatial transcriptomics data at the molecule level, and to have standardised data formats accross vendors.
Peter Hickey (23:43:04) (in thread): > will ask our team if we have any that we can share easily
Shila Ghazanfar (23:46:02) (in thread): > thanks so much pete! what i’m seeing so far is that the cell segmentation component might be different to what’s on the showcase data on vizgen website
Peter Hickey (23:46:45) (in thread): > how annoying!
Peter Hickey (23:47:06) (in thread): > do you need the data itself or would the output oftree(showing the directory structure) tell you what you need to know?
Shila Ghazanfar (23:48:14) (in thread): > output of tree would be extremely helpful!! of course some data to test would be extra helpful too but i understand the data is so new and most likely not able to share just yet
Peter Hickey (23:48:44) (in thread): > i’ve asked and will let you know when i get an answer :)
Shila Ghazanfar (23:48:55) (in thread): > thanks so much!
2023-06-26
Ludwig Geistlinger (09:10:45) (in thread): > Hi Shila, have you seen this minimal output specification from:https://vizgen.com/products/(Section “The MERSCOPE Workflow”?):
Ludwig Geistlinger (09:11:23) (in thread): - File (PNG): Screenshot 2023-06-26 at 3.08.56 PM.png
Ludwig Geistlinger (09:13:05) (in thread): > It seems you have taken a look at the three example datasets underhttps://vizgen.com/data-release-program/? What kind of differences have you observed to what’s coming off the machine? I can ask Jeff.
Shila Ghazanfar (18:44:24) (in thread): > thanks ludwig! i havent seen any data off of machines (yet!:smile:) so just looking to confirm this is the case — with thanks also to@Peter Hickeyit appears vizgen provides cell boundaries as a folder containing N(=number of FOVs(?)) .HDF5 files, this is an interesting choice to me as I’ve been seeing large .tif masks or tabulated .csv with polygon vertices. I suppose this ensures that the boundaries are stored with each z-slice potentially being different..
Peter Hickey (18:46:24) (in thread): > I’ll re-post what i shared with Shila to the channel
Peter Hickey (18:47:13): > This is a training sample we ran in August, 2022. > Ray is the person who runs our MERSCOPE. - Attachment: Attachment > Ray tells me this is the “raw data (as in raw images)” > % tree -d merfish_raw_data/202208261120_20220826MouseBrainTraining_VMSC04501/ > merfish_raw_data/202208261120_20220826MouseBrainTraining_VMSC04501/ > ├── data > ├── low_resolution > │ └── LowResolutionMosaic > │ ├── Data > │ ├── LowResolutionMosaic > │ │ ├── GenerateMosaic > │ │ │ ├── images > │ │ │ ├── log > │ │ │ └── tasks > │ │ ├── NoneWarp > │ │ │ ├── log > │ │ │ └── tasks > │ │ └── SimpleGlobalAlignment > │ │ ├── log > │ │ └── tasks > │ └── settings > ├── regions_images > ├── seg_preview > │ ├── analysis > │ │ └── seg_preview > │ │ ├── NoneWarp > │ │ │ ├── log > │ │ │ └── tasks > │ │ ├── SimpleGlobalAlignment > │ │ │ ├── log > │ │ │ └── tasks > │ │ ├── snakemake > │ │ └── WatershedSegment_PolyT > │ │ ├── features > │ │ ├── log > │ │ └── tasks > │ ├── data > │ └── settings > └── settings > > 34 directories
Peter Hickey (18:47:23): - Attachment: Attachment > And that this is the “decoded images and ready-to-use data (i.e. cell by gene matrix, vzg visualiser file)” > % tree -d merfish_output/202208261120_20220826MouseBrainTraining_VMSC04501/ > merfish_output/202208261120_20220826MouseBrainTraining_VMSC04501/ > └── region_0 > ├── cell_boundaries > └── images > > 3 directories
Peter Hickey (18:48:51): > And contents of ofcell_boundariesdirectory … 1,476 HDF5 files > > % ll merfish_output/202208261120_20220826MouseBrainTraining_VMSC04501/region_0/cell_boundaries/ | head > total 0 > -rwxrwx--- 1 yip.r microscopy 989K Aug 27 2022 feature_data_0.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.0M Aug 27 2022 feature_data_1000.hdf5 > -rwxrwx--- 1 yip.r microscopy 2.5M Aug 27 2022 feature_data_1001.hdf5 > -rwxrwx--- 1 yip.r microscopy 1.8M Aug 27 2022 feature_data_1002.hdf5 > -rwxrwx--- 1 yip.r microscopy 650K Aug 27 2022 feature_data_1003.hdf5 > -rwxrwx--- 1 yip.r microscopy 881K Aug 27 2022 feature_data_1004.hdf5 > -rwxrwx--- 1 yip.r microscopy 908K Aug 27 2022 feature_data_1005.hdf5 > -rwxrwx--- 1 yip.r microscopy 1.8M Aug 27 2022 feature_data_1006.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.2M Aug 27 2022 feature_data_1007.hdf5 > (base) vc7-shared 527 % ll merfish_output/202208261120_20220826MouseBrainTraining_VMSC04501/region_0/cell_boundaries/ | tail > -rwxrwx--- 1 yip.r microscopy 4.0M Aug 27 2022 feature_data_992.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.7M Aug 27 2022 feature_data_993.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.4M Aug 27 2022 feature_data_994.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.9M Aug 27 2022 feature_data_995.hdf5 > -rwxrwx--- 1 yip.r microscopy 2.9M Aug 27 2022 feature_data_996.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.3M Aug 27 2022 feature_data_997.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.0M Aug 27 2022 feature_data_998.hdf5 > -rwxrwx--- 1 yip.r microscopy 3.3M Aug 27 2022 feature_data_999.hdf5 > -rwxrwx--- 1 yip.r microscopy 2.5M Aug 27 2022 feature_data_99.hdf5 > -rwxrwx--- 1 yip.r microscopy 1.8M Aug 27 2022 feature_data_9.hdf5 >
2023-06-28
Bárbara Zita Peters Couto (07:46:13): > @Bárbara Zita Peters Couto has joined the channel
2023-07-28
Konstantinos Daniilidis (13:47:25): > @Konstantinos Daniilidis has joined the channel
Benjamin Yang (15:59:02): > @Benjamin Yang has joined the channel
2023-08-02
Jamin Liu (14:43:52): > @Jamin Liu has joined the channel
2023-08-04
Trisha Timpug (09:36:14): > @Trisha Timpug has joined the channel
2023-08-07
Jiaji George Chen (11:23:29): > @Jiaji George Chen has joined the channel
2023-08-09
Sean Davis (06:42:11): > My university just purchased a CosMx. My understanding is that the data are processed on a cloud-based platform. Does anyone have any comment on the required network bandwidth for the processing and accessing the resulting data? It looks to me like the data are large (about 400GB per sample, raw data, extracted). [Parenthetically, this is the first time I have heard of someone purchasing insurance, not against machine breakage, but against legal action!]
2023-08-11
Shila Ghazanfar (03:43:01) (in thread): > Hi Sean! I don’t know how useful this is, but my university is going through the same thing.. from the talks i’ve heard from NanoString they really want processing and most/all low-level analysis be done on the cloud atomx platform ie you upload your scripts not vice versa. as for the bandwidth to pull down the data i’m not so sure, curious about others’ experiences..
2023-08-14
Leonardo Collado Torres (13:14:02): > https://github.com/Bioconductor/BiocFileCache/issues/48 - Attachment: #48 curl error with bfcrpath: HTTP/2 stream 1 was not closed cleanly: PROTOCOL_ERROR (err 1) > Hi BiocFileCache team, > > Last week at BioC2023 I ran into some issues with spatialLIBD::fetch_data("spatialDLPFC_Visium_modeling_results") and related functions after I wiped out clean my BiocFileCache on my laptop and forced it to re-download data. I made a reprex below abstracting away the spatialLIBD layer. > > > library("BiocFileCache") > #> Loading required package: dbplyr > bfc <- BiocFileCache::BiocFileCache() > BiocFileCache::bfcrpath(bfc, "http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1") > #> adding rname '[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)' > #> Warning in value[[3L]](cond): > #> trying to add rname '[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)' produced error: > #> HTTP/2 stream 1 was not closed cleanly: PROTOCOL_ERROR (err 1) > #> Error in BiocFileCache::bfcrpath(bfc, "[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)"): not all 'rnames' found or unique. > traceback() > #> No traceback available > options(width = 120) > sessioninfo::session_info() > #> ─ Session info ─────────────────────────────────────────────────────────────────────────────────────────────────────── > #> setting value > #> version R version 4.3.1 (2023-06-16) > #> os macOS Ventura 13.4 > #> system aarch64, darwin20 > #> ui X11 > #> language (EN) > #> collate en_US.UTF-8 > #> ctype en_US.UTF-8 > #> tz America/New_York > #> date 2023-08-11 > #> pandoc 3.1.5 @ /opt/homebrew/bin/ (via rmarkdown) > #> > #> ─ Packages ─────────────────────────────────────────────────────────────────────────────────────────────────────────── > #> package * version date (UTC) lib source > #> BiocFileCache * 2.8.0 2023-04-25 [1] Bioconductor > #> bit 4.0.5 2022-11-15 [1] CRAN (R 4.3.0) > #> bit64 4.0.5 2020-08-30 [1] CRAN (R 4.3.0) > #> blob 1.2.4 2023-03-17 [1] CRAN (R 4.3.0) > #> cachem 1.0.8 2023-05-01 [1] CRAN (R 4.3.0) > #> cli 3.6.1 2023-03-23 [1] CRAN (R 4.3.0) > #> curl 5.0.1 2023-06-07 [1] CRAN (R 4.3.0) > #> DBI 1.1.3 2022-06-18 [1] CRAN (R 4.3.0) > #> dbplyr * 2.3.3 2023-07-07 [1] CRAN (R 4.3.0) > #> digest 0.6.33 2023-07-07 [1] CRAN (R 4.3.0) > #> dplyr 1.1.2 2023-04-20 [1] CRAN (R 4.3.0) > #> evaluate 0.21 2023-05-05 [1] CRAN (R 4.3.0) > #> fansi 1.0.4 2023-01-22 [1] CRAN (R 4.3.0) > #> fastmap 1.1.1 2023-02-24 [1] CRAN (R 4.3.0) > #> filelock 1.0.2 2018-10-05 [1] CRAN (R 4.3.0) > #> fs 1.6.3 2023-07-20 [1] CRAN (R 4.3.0) > #> generics 0.1.3 2022-07-05 [1] CRAN (R 4.3.0) > #> glue 1.6.2 2022-02-24 [1] CRAN (R 4.3.0) > #> htmltools 0.5.5 2023-03-23 [1] CRAN (R 4.3.0) > #> httr 1.4.6 2023-05-08 [1] CRAN (R 4.3.0) > #> knitr 1.43 2023-05-25 [1] CRAN (R 4.3.0) > #> lifecycle 1.0.3 2022-10-07 [1] CRAN (R 4.3.0) > #> magrittr 2.0.3 2022-03-30 [1] CRAN (R 4.3.0) > #> memoise 2.0.1 2021-11-26 [1] CRAN (R 4.3.0) > #> pillar 1.9.0 2023-03-22 [1] CRAN (R 4.3.0) > #> pkgconfig 2.0.3 2019-09-22 [1] CRAN (R 4.3.0) > #> purrr 1.0.1 2023-01-10 [1] CRAN (R 4.3.0) > #> R.cache 0.16.0 2022-07-21 [1] CRAN (R 4.3.0) > #> R.methodsS3 1.8.2 2022-06-13 [1] CRAN (R 4.3.0) > #> R.oo 1.25.0 2022-06-12 [1] CRAN (R 4.3.0) > #> R.utils 2.12.2 2022-11-11 [1] CRAN (R 4.3.0) > #> R6 2.5.1 2021-08-19 [1] CRAN (R 4.3.0) > #> reprex 2.0.2 2022-08-17 [1] CRAN (R 4.3.0) > #> rlang 1.1.1 2023-04-28 [1] CRAN (R 4.3.0) > #> rmarkdown 2.23 2023-07-01 [1] CRAN (R 4.3.0) > #> RSQLite 2.3.1 2023-04-03 [1] CRAN (R 4.3.0) > #> rstudioapi 0.15.0 2023-07-07 [1] CRAN (R 4.3.0) > #> sessioninfo 1.2.2 2021-12-06 [1] CRAN (R 4.3.0) > #> styler 1.10.1 2023-06-05 [1] CRAN (R 4.3.0) > #> tibble 3.2.1 2023-03-20 [1] CRAN (R 4.3.0) > #> tidyselect 1.2.0 2022-10-10 [1] CRAN (R 4.3.0) > #> utf8 1.2.3 2023-01-31 [1] CRAN (R 4.3.0) > #> vctrs 0.6.3 2023-06-14 [1] CRAN (R 4.3.0) > #> withr 2.5.0 2022-03-03 [1] CRAN (R 4.3.0) > #> xfun 0.39 2023-04-20 [1] CRAN (R 4.3.0) > #> yaml 2.3.7 2023-01-23 [1] CRAN (R 4.3.0) > #> > #> [1] /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/library > #> > #> ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── > > > Created on 2023-08-11 with reprex v2.0.2 > > I was surprised by this as everything is working ok at GitHub actions using the biocthis workflow as well as in the Bioconductor BBS system. > > In the 30 min prior to our spatialLIBD BioC2023 demo with @lahuuki, I tried resolving this. By googling the error message, I realized that curl (and httr), BiocFileCache dependencies, now use http version 2 instead of 1.1. See jeroen/curl#232 for a related issue. There they link to https://cran.r-project.org/web/packages/curl/vignettes/intro.html#Disabling_HTTP2|https://cran.r-project.org/web/packages/curl/vignettes/intro.html#Disabling_HTTP2 for disabling HTTP2, but I’m not sure if I can do that as a BiocFileCache user. Maybe through the config argument https://github.com/search?q=repo%3ABioconductor%2FBiocFileCache%20config&type=code. But that doesn’t seem to be the case (maybe I’m not specifying config where I should): > > > > BiocFileCache::bfcrpath(bfc, "[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)", config = list(http_version = 2)) > adding rname '[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)' > |==================================================================================| 100% > > Error in BiocFileCache::bfcrpath(bfc, "[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)", : > not all 'rnames' found or unique. > In addition: Warning message: > In value[[3L]](cond) : > trying to add rname '[http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1](http://www.dropbox.com/s/srkb2ife75px2yz/modeling_results_BayesSpace_k09.Rdata?dl=1)' produced error: > HTTP/2 stream 1 was not closed cleanly: PROTOCOL_ERROR (err 1) > > > What I find weird is that the download does actually happen, but at the very end it errors out. So I’m not 100% sure that the error is related to the http version 1.1 vs 2 change. > > Hm… maybe something else changed in BiocFileCache::bfcrpath() that I’m not aware of, but well, in that case, I would expect things to break on GitHub Actions and BBS as well. > > 1 to 3 students at https://github.com/lcolladotor/cshl_rstats_genome_scale_2023 ran into this error and I thought it was something related their laptop configurations, and for the purpose of that course, we could get around things by having them download the data from Dropbox manually. > > Thanks,
> Leo
Leonardo Collado Torres (13:14:24): > TL DR, if you are having issues downloading data on macOS, you likely need to update your operating system.
Leonardo Collado Torres (13:14:44): > This is the root of the issue I ran into at#bioc-conference-everyonefor ourspatialLIBDdemo with@Louise Huuki
Louise Huuki (13:14:59): > @Louise Huuki has joined the channel
Vince Carey (15:45:46) (in thread): > hi lots of travel going on now will address later in the week
2023-08-31
Frederick Tan (09:10:44): > Any experiences with the Curio Seeker kit (curiobioscience.com), good or bad?
Jenny Drnevich (10:46:59) (in thread): > Not directly yet. We had someone here very interested in using it and our Cyto facility got into heavy discussions with Curio and us about it… which went nowhere. Mainly because the person interested wanted to do plant tissue and Curio kept saying they would have recommendations for plant in two weeks (that was 6 months ago). I was supposed to try out their software with practice data, but could only get it through very bizarre ways - a time-sensitive link to a tarball and a vague mention of “instructions are in the README”. You can’t even see any real documentation on it now without giving them your information. I was too busy to jump through the hoops to try out their software for a project that might not happen because they can’t support plants. Oh, IIRC you don’t even take an actual image of the tissue slice and it seemed like all spatial aspects were done by expression levels only…
Frederick Tan (10:58:26) (in thread): > Sounds like they may be taking a page out of ONT?:slightly_smiling_face:Thanks for sharing your experience. We’re meeting with their rep in a couple of weeks (who used to be our 10x Genomics rep).
Jenny Drnevich (11:01:17) (in thread): > Good luck! It seems like it could be a good technology for mammalian tissues with well defined morphologies that fit within the scale - like mouse brains!
Frederick Tan (11:16:23) (in thread): > Hoping it can work for zebrafish, drosophila, and corals:crossed_fingers:
2023-09-01
Eddie Ruiz (09:33:02): > @Eddie Ruiz has joined the channel
2023-09-12
Jiwoon Park (21:49:15): > @Jiwoon Park has joined the channel
2023-09-13
Ruben Dries (10:57:26): > @Ruben Dries has joined the channel
Christopher Chin (17:05:11): > @Christopher Chin has joined the channel
Joselyn Chávez (21:07:56): > @Joselyn Chávez has joined the channel
2023-09-14
Leonardo Collado Torres (21:49:29): > https://x.com/spencef/status/1702360289306415249?s=12&t=fPqNoppoJ109s623S-JHFw - Attachment (X (formerly Twitter)): spence fast :dna: on X > Hey single cell friends. Amazing new update to 10xGenomics Loupe browser (Version 7.0)! All new look and even more intuiative. New multisample DE. Also, now reimport your Seurat objects back into loupe with LoupeR! :dna::microscope:https://t.co/NBF1FdPshg
Leonardo Collado Torres (21:49:46): > multi-sample support on Loupe:thinking_face:
Jenny Drnevich (23:15:11) (in thread): > Yay! I was just asking our 10X rep when we were going to be able to import properly-normalized data into Loupe for clients. I’ll have to play around with it…
Jenny Drnevich (23:17:10) (in thread): > (the funny thing is that I just asked her the day before this was released, and she said she’d create a ticket for it, but “they don’t always tell us field reps what is being worked on or when it will be released!)
2023-09-15
Leo Lahti (04:56:34): > @Leo Lahti has joined the channel
2023-09-19
Lambda Moses (21:35:41): > TheSpatialExperimentobject has the specialsample_idcolumn incolDatabecause different samples can have overlapping coordinate values. Spatial transcriptomics datasets often have multiple samples, from different subjects, case/control conditions, and sometimes serial sections or different locations from the same subject. It might make sense to compute say Moran’s I jointly across different samples in the same treatment group in addition to individually for each tissue section, so Moran’s I can be compared across samples and conditions as part of ESDA. Within one sample, there’re global spatial statistics that gives one result for the entire dataset and local spatial statistics giving a result for each spatial location. But here with multiple samples, we have a spectrum from global to local for spatial statistics: the entire study with all samples, all samples from the same case/control condition, samples from the same subject, samples from the same piece of tissue from the same subject, serial sections (images can be registered to make 3D data), regions within each section, and cells/spots within each section. > > While I’m considering expandingsample_idto this hierarchy in a future version ofSpatialFeatureExperiment, I want to ask if something like this already exists, because the hierarchy of study design and samples isn’t new to spatial. Has anyone heard of an existing package with a data structure managing this hierarchy?
2023-09-20
Alik Huseynov (04:22:00): > @Alik Huseynov has joined the channel
Vince Carey (06:37:30): > @Lambda MosesI am not sure if this addresses your question at all, or if it is too specialized, but nlme has a groupedData class that could be relevant. There is a formula idiom for hierarchical structure specification.
Helena L. Crowell (08:07:10) (in thread): > Not sure exactly where this is going, butTreeSummarizedExperimentcomes to mind (storing a hierarchical structure for the cells, which the above kinda sounds like).
Lambda Moses (15:16:34) (in thread): > Sounds like it’s relevant. Thanks!
Lambda Moses (15:18:31) (in thread): > Thanks, could be relevant
Leonardo Collado Torres (21:58:40) (in thread): > ohh
Leonardo Collado Torres (21:58:51) (in thread): > yeah, I guess it’s hard to keep everyone on the same page on a large company
Leonardo Collado Torres (21:59:03) (in thread): > or well, across teams in any company
2023-09-25
Kerim Secener (09:12:49): > @Kerim Secener has joined the channel
2023-09-29
Wes W (10:46:58): > @Wes W has joined the channel
2023-10-18
Ludwig Geistlinger (12:07:24): > Our next seminar on Mon, Oct 23, 10-11 AM EDT, will feature Luca Marconato who will present anew frameworkfor the analysis of spatial omics data. Join us under the zoom link provided in the flyer below! - File (PDF): CCB_Seminar_SpatialData.pdf
2023-11-02
Sunil Poudel (10:51:51): > @Sunil Poudel has joined the channel
2023-11-13
Estella Dong (13:09:35): > @Estella Dong has joined the channel
2023-11-16
Sue McClatchy (09:59:05): > @Sue McClatchy has joined the channel
2023-12-13
Paul Myers (09:45:18): > @Paul Myers has joined the channel
2023-12-14
Marc Elosua (15:39:45): > @Marc Elosua has joined the channel
2024-01-04
Artür Manukyan (05:43:42): > @Artür Manukyan has joined the channel
2024-01-08
Connie Li Wai Suen (17:07:17): > @Connie Li Wai Suen has joined the channel
2024-01-10
Bernie Mulvey (15:04:08): > @Bernie Mulvey has joined the channel
Alik Huseynov (15:36:59): > https://www.10xgenomics.com/products/visium-hd-spatial-gene-expression?utm_medium=emai[…]zJ8O-hPHS1j0Im8kabNepAfhsB1o_Avmac35s7VWIBEarQyUOwdKN-dRlsQ - Attachment (10x Genomics): Visium HD Spatial Gene Expression - 10x Genomics > A whole new visual canvas to explore single cell data
Artür Manukyan (15:40:44) (in thread): > also thishttps://twitter.com/nanostringtech/status/1745083138542313518?t=qrzuob9Pl-_jlKC_mNPHUA&s=19 - Attachment (X (formerly Twitter)): NanoString (@nanostringtech) on X > NanoString CEO Brad Gray reveals spatial biology moonshot at #JPM2024! Unveiling the CosMx SMI Whole Transcriptome Panel— the first true single-cell imaging of >18,000 genes. Join us today at 5:15 PM PST :globe_with_meridians: https://t.co/ezSijOeTVr.
Jenny Drnevich (16:35:08) (in thread): > Or this pre-print for Open-ST that by-passes Visium/CosMx et al. by directly using an Illumina flowcell to do the spatial capture. It is generating a lot of interest in our sequencing center
Artür Manukyan (16:57:23) (in thread): > Amazing work by upstairs people here in MDC Berlin
Lambda Moses (17:10:27) (in thread): > OK, so I need to play around with SpatialData and an example dataset, or maybe just terra, and if it goes well I should make Voayger work with SpatialData. This is obviously raster.
2024-01-11
Nilesh Kumar (12:01:05): > @Nilesh Kumar has joined the channel
2024-01-12
Boyi Guo (11:11:32) (in thread): > VisiumHD doesn’t seem have an example data, or exact timeline of releasing it.https://x.com/10xGenomics/status/1745501459826008298?s=20 - Attachment (X (formerly Twitter)): 10x Genomics (@10xGenomics) on X > @BoyiGuo Datasets should be available around when the product begins shipping, stay tuned! In the meantime you can learn more about Visium HD here: https://t.co/qtN6wJPPf7
Alik Huseynov (11:19:17) (in thread): > Thanks@Boyi Guo, I was wondering as well, hopefully one can get the toy dataset soon. > Anyone know when they start shipping?
2024-02-02
Lambda Moses (19:55:06): > I invite you to the Voyager workshop and hackathon on March 4-8 at Caltech, Pasadena, CA:https://fosstodon.org/@lambdamoses/111864733704412487 - Attachment (Fosstodon): Lambda Moses (@lambdamoses@fosstodon.org) > Attached: 1 image > > Interested in spatial -omics data analysis? Interested in getting started with R package dev? Or are you interested in using Voyager (https://pachterlab.github.io/voyager/) and need new features? Finding bugs in Voyager? Please come to the Voyager workshop and hackathon on March 4-8 at Caltech! More info & RSVP here: https://cryptpad.fr/form/#/2/form/view/vcSr+2tQw5UPOP3M5FIOxyL3jV8NP72YOezM74cpCfY/ @bioconductor@genomic.social #spatial #rstats #LosAngeles #Bioinformatics
2024-02-07
Davide Risso (02:16:23) (in thread): > They told me March…at least here in Europe
Jenny Drnevich (09:03:08) (in thread): > We are co-hosting a workshop with 10X here on Feb 29, mainly on Xenium but also Visium HD - I’m hoping they will be able to give us more details then.
2024-03-05
Pratibha Panwar (01:20:55): > @Pratibha Panwar has joined the channel
2024-03-06
Michael Lynch (08:00:20): > @Michael Lynch has joined the channel
2024-03-15
Davide Risso (08:01:31): > Hi@Hervé Pagès, I’m adding you to this channel as last Wednesday@Vince Carey@Helena L. Crowell@Ludwig Geistlingerand I discussed the needs for facilitating spatial transcriptomics in Bioconductor. Two major hurdles were highlighted: a better support for Zarr and an S4 object that would help coordinate image transformations and similar operations across different layers (images, shapes, tables), similar to what SpatialData achieves in Python.@Vince Careysuggested to point you to an example of public data and I thought that it would be good to have this discussion in this channel so that if other people have input/feedback they can participate. > > So, this is a link to a tutorial that shows how to align different images related to the same sample using SpatialData in Python:https://spatialdata.scverse.org/en/latest/tutorials/notebooks/notebooks/examples/alignment_using_landmarks.htmlThere, you will see the links to the public data. > > Another typical operation that we would want to have is integration/aggregation of signal across different layers, as done here:https://spatialdata.scverse.org/en/latest/tutorials/notebooks/notebooks/examples/aggregation.htmlFinally,@Helena L. Crowellhas started a first implementation of these concepts here:https://github.com/HelenaLC/SpatialDataIt uses basilisk and zellkonverter IIRC but we also discussed that a more native R implementation could have its advantages (based on rarr, geoparquet, and the like). It may also be good to have a look at the design document of SpatialData:https://spatialdata.scverse.org/en/latest/design_doc.htmlHappy to discuss this further either here or through another group zoom meeting. (Sorry for the long message!)
Hervé Pagès (08:01:34): > @Hervé Pagès has joined the channel
Artür Manukyan (08:05:22) (in thread): > There is a channel dedicated to Helena’s SpatialData R implementation by the way, here:https://community-bioc.slack.com/archives/C0558SHL814
Davide Risso (08:09:34) (in thread): > Thanks!Didn’tknow about that channel!
2024-03-18
Nick Eagles (10:03:51): > @Nick Eagles has joined the channel
Luca Marconato (11:12:02) (in thread): > Sounds great! I will give a pass to the design doc and check if it’s up-to-date.~It will be updated by tomorrow.~Updated!:blush:
Hervé Pagès (14:05:05) (in thread): > Sounds like an interesting project. Happy to get involved, even though my plate is full at the moment so won’t be able to do much before the 3.19 release.
2024-03-21
Davide Risso (04:41:59): > I guess it was good timing for my previous message…Congrats@Luca Marconato@Giovanni Palla@Isaac Virshup
- Attachment (Nature): SpatialData: an open and universal data framework for spatial omics > Nature Methods - SpatialData is a user-friendly computational framework for exploring, analyzing, annotating, aligning and storing spatial omics data that can seamlessly handle large multimodal…
Luca Marconato (10:29:53) (in thread): > Thanks a lot!:blush:Looking forward for the next developments!
2024-03-26
Aaron Lun (20:42:09): > thoughts of deprecation ofasSparse=in*Pairsfunctions of the SCE:https://github.com/drisso/SingleCellExperiment/issues/73#issuecomment-2019477191; comments welcome. - Attachment: Comment on #73 colPairs(x, asSparse = TRUE) fails when there are no metadata columns > I’ll take the PR since it’s already been put together, but longer term, I am wondering whether it would be better to (i) deprecate the asSparse= and (ii) export the .hits2mat function (and its counterpart for converting a matrix to a Hits object). This would simplify the interface; the SCE only cares about Hits going in and out, and exactly how this is converted to/from a sparse matrix is up to the user/application outside of the SCE. > > I don’t ever use the colPairs and rowPairs functionality so I have no relevant experience/opinion on this matter, but happy to take thoughts from people who do use it.
2024-03-27
abhich (05:46:01): > @abhich has joined the channel
2024-04-04
Tung Trinh (23:38:53): > @Tung Trinh has joined the channel
2024-04-06
Alik Huseynov (06:21:08): > Our ST techs comparison study (Visium, RNAScope, Vizgen, Xenium and Molecular Cartography, as well as snRNAseq) is finally online.https://www.biorxiv.org/content/10.1101/2024.04.03.586404v1 - Attachment (bioRxiv): Comparison of spatial transcriptomics technologies used for tumor cryosections > Background: Spatial transcriptomics (ST) methods provide single cell transcriptome profiles within the endogenous cell tissue context. Thus, they are ideally suited to resolve intra-tumor heterogeneity and interactions between tumor and non-malignant cells in their microenvironment. However, different ST technologies exist and are rapidly evolving. It is frequently not clear how they address a given research question in cancer research. Here, we investigated fresh frozen tissue sections from medulloblastoma with enhanced nodularity (MBEN) using four distinct imaging-based ST methodologies (RNAscope HiPlex, Molecular Cartography, MERFISH/Merscope, and Xenium), alongside sequencing-based ST on Visium slides. Results: Our comparative case study describes critical aspects of the different workflows and identifies informative parameters to assess the experimental results. We evaluate sensitivity and specificity across platforms, show how technology dependent features affect the results and provide guidance for the experimental design. Furthermore, we demonstrate how cell segmentation can be improved and/or additional custom readouts can be integrated by reimaging of the slides after the ST experiment. Conclusions: The different ST methods come with specific features that should be considered in the method selection and experimental design. We anticipate that the insight gained in our study will facilitate successful application of ST for the analysis of fresh frozen tissue sections of solid tumors. > > ### Competing Interest Statement > > The authors have declared no competing interest.
2024-04-18
Philipp Sergeev (03:01:57): > @Philipp Sergeev has joined the channel
2024-04-29
Martin Emons (02:24:38): > @Martin Emons has joined the channel
Samuel Gunz (09:34:53): > @Samuel Gunz has joined the channel
Jacqui Thompson (19:00:27): > @Jacqui Thompson has joined the channel
2024-04-30
Jenny Drnevich (12:13:23) (in thread): > Visium HD datasets are now available - Attachment (10x Genomics): Datasets - 10x Genomics
Boyi Guo (21:06:08) (in thread): > Yep, very excited seeing it.https://x.com/BoyiGuo/status/1777884443534889403I was just trying to understand the file structure of thespacerangeroutput and see if it is possible to load it to some bioc infrastructure. So far, the biggest burden seems to be the outputs folder uses a new file format .parquet that replaces tissue_position.csv comparing to a Visiumspacerangeroutput. > > A lab mate of mine triedSeuratwhich is capable of loading it. - Attachment (X (formerly Twitter)): Boyi Guo 郭博逸 (@BoyiGuo) on X > Comparing to #Visium, @10xGenomics adds a new h5 file, Feature Slice H5, as the Spaceranger 3.0 output for #VisiumHD > > A python example is provided how to bin 2um at customizable resolution. > > https://t.co/AcOuNl4eZ6
Lambda Moses (21:18:21) (in thread): > We’re working on updating the SFE package to load it. It’s not a big edit from the function to read old fashioned Visium.
Boyi Guo (21:22:42) (in thread): > I agree. Look forward to it! > > I’m also just curious that if you have specific strategy to managing the outputs with different resolution? As a user, I’m concerned about file size and overloading some part of data that is useless for downstream. > > Specifically imaging people in near future could store multiple VisiumHD samples in the same structure.
Lambda Moses (21:24:41) (in thread): > For the gene count matrix we already have DelayedArray. Also there are implications to doing the analyses at different resolution so we wouldn’t always use the 2 um resolution. For the geometries, which can take up a lot of RAM, I’m considering apache sedona or DuckDB’s spatial extension which support geometric operations out of memory.
Lambda Moses (21:25:16) (in thread): > I also need to play around with VisiumHD data a bit more to decide what to do.
2024-05-01
Luca Marconato (15:27:50) (in thread): > From the Python side, here is the reader that we developed:https://github.com/scverse/spatialdata-io/blob/main/src/spatialdata_io/readers/visium_hd.pyand you can find and example notebook here:https://spatialdata.scverse.org/en/latest/tutorials/notebooks/notebooks/examples/technology_visium_hd.html. > > We are working on improving the performance when plotting the high-res datasets (as we represent it as vector data), and also we are working on a raster representation of the data.
2024-05-03
Benedetta Banzi (04:53:40): > @Benedetta Banzi has joined the channel
2024-05-04
Aedin Culhane (12:49:46): > upcoming online course on Spatial Omics with R/Bioconductor, taking place in just two weeks.:https://www.physalia-courses.org/courses-workshops/spatial-transcriptomics/ - Attachment (physalia-courses): spatial Omics in R/Bioconductor > 20-22 May 2024 To foster international participation, this course will be held online
2024-05-05
Carlos Enrique Balcazar Lopez (02:07:03): > @Carlos Enrique Balcazar Lopez has joined the channel
2024-05-06
Michal Kolář (11:56:15): > @Michal Kolář has joined the channel
Matt Stone (13:37:03): > @Matt Stone has joined the channel
2024-05-07
Carlo Pecoraro (12:10:53): > @Carlo Pecoraro has joined the channel
2024-05-15
Sunil Nahata (08:30:54): > @Sunil Nahata has joined the channel
2024-06-07
Alik Huseynov (08:08:45) (in thread): > FYI,SpatialFeatureExperiment::read10xVisiumSFEnow supports loading VisiumHD. > > devtools::install_github("pachterlab/SpatialFeatureExperiment", ref = "devel") >
2024-06-10
Alik Huseynov (10:36:54): > Update onSpatialFeatureExperimentandVoyager: > * now supports VisiumHD, optimized readers for Vizgen/MERSCOPE and Xenium (including Xenium Prime 5K (XOA v3) multimodal cell segmentation), > * operations on transcript (molecule) coordinates -rowGeometries > * loading spatial tech-specific image data (including OME-TIFF) > * conversion from Seurat object(currently Visium, VisiumHD, Vizgen, Xenium) to SFE for exploratory spatial data analysis using Voyager. > > * .Rmdvignette > These packages are currently under active development, so, much more to come soon. > If anyone have issues or questions, simply DM us@Lambda Moses@Alik Huseynovor open a package-specific issue for SFE or Voyager.
2024-06-11
Niklas Beumer (01:52:37): > @Niklas Beumer has joined the channel
Ziru Chen (04:36:20): > @Ziru Chen has joined the channel
Ivan Osinnii (08:56:44): > @Ivan Osinnii has joined the channel
2024-06-14
Joselyn Chávez (10:02:30): > Hi all, we are hosting a 3-day online workshop for doing spatial analysis with the Giotto package.Join us! Feel free to reach out if you have any questions - File (PDF): giotto_flyer
Jenny Drnevich (10:04:09) (in thread): > This looks amazing! Is there any fee for the workshop?
Joselyn Chávez (10:06:29) (in thread): > @Jenny Drnevichno fees, is a free event
Jenny Drnevich (10:07:03) (in thread): > :star-struck::tada:
2024-06-21
Alik Huseynov (03:13:47): > EU spatial conference in Berlinhttps://spatialbiologysociety.eu/ - Attachment (European Society for Spatial Biology): Homepage - European Society for Spatial Biology > Inaugural conference of the European Society for Spatial Biology (ESSB) Register for the conference 12 – 13 December 2024, Berlin, Germany Venue: Hotel & Conference Center Titanic Chaussee Berlin Welcome On December 12 and 13, 2024, the inaugural conference of the ESSB will take place here in Berlin. We expect an interdisciplinary audience from […]
2024-06-26
Bianka Berki (06:01:01): > @Bianka Berki has joined the channel
2024-06-30
Nicolas Peterson (13:09:11): > @Nicolas Peterson has joined the channel
2024-07-05
Antonin Thiébaut (04:37:35): > @Antonin Thiébaut has joined the channel
2024-07-11
Hothri Moka (07:20:07): > @Hothri Moka has joined the channel
2024-07-31
Zahraa W Alsafwani (17:21:40): > @Zahraa W Alsafwani has joined the channel
2024-08-19
Rema Gesaka (09:41:37): > @Rema Gesaka has joined the channel
2024-08-24
Sean Davis (09:31:51): > https://datascience.cancer.gov/news-events/news/solve-challenges-spatial-omics-data-human-tumor-atlas-network-htan-data-jamboree - Attachment (datascience.cancer.gov): Solve Challenges with Spatial Omics Data at the Human Tumor Atlas Network (HTAN) Data Jamboree 2024 | CBIIT > Are you interested in using spatial omics and single-cell approaches in your cancer research? Participate in the HTAN Data Jamboree, where you’ll work with a team to build unique solutions that solve problems in cancer research. Submit your application by September 6.
2024-08-27
NILESH KUMAR (11:36:40): > @NILESH KUMAR has joined the channel
2024-09-03
Vince Carey (16:36:57): > is thishttps://vjcitn.shinyapps.io/XenLUAD/redundant with other functionality in bioc ? it takes a little while to warm up. a 5k panel seems too big to run in 8GB inshinyapps.io, with current design….
2024-09-04
Luca Marconato (11:28:31) (in thread): > > We are working on improving the performance when plotting the high-res datasets (as we represent it as vector data), and also we are working on a raster representation of the data. > This is now available with the functionrasterize_bins()(developed together with Quentin Blampey). > > The new API: > * operates lazily and channel-wise (otherwise it would require 180GB to store the rasterized the data) > * adjusts for pixel sizes/coordinate transformations. > You can find thecode here, and examplenotebook here, and in 2 weeks we are discussing the functionality, and a few other Visium HD related use cases, inthis 15 min webinar. - Attachment (X (formerly Twitter)): 10x Genomics (@10xGenomics) on X > Your reaction to #VisiumHD data analysis may be somewhere between :nerd_face: and :dizzy_face:. Wherever you land, join this webinar to see how intuitive and adaptable high-def #spatialtranscriptomics data exploration can be with tools like Bin2Cell & SpatialData. https://t.co/ARrNB93Skk
2024-09-06
Estella Dong (08:16:40): > @Marcel Ramos Pérez@Dario RighelliWe are trying to wrap up everything that came out from the EuroBioc2024, and we are going to create a to-do list. Anyone interested can join us and contribute. Marcel will start a Github Project board for this working group.:slightly_smiling_face:
Alik Huseynov (08:50:29) (in thread): > good idea! I would be interested to take a look at todo list and contribute where I can.
2024-09-07
Marcel Ramos Pérez (03:23:02) (in thread): > Thank you Estella! Here is the link to the project board.https://github.com/orgs/Bioconductor/projects/16cc:@Vince Carey
2024-09-09
Dario Righelli (08:26:03) (in thread): > @Marcel Ramos PérezIs there a way for me to add some tasks to the todo list?
Alik Huseynov (08:32:00) (in thread): > right now it’s just Visium-related things on the todo list. I guess when that list has more items, then people could claim and work on them.
Artür Manukyan (08:32:22) (in thread): > are you guys planning to hold periodic meetings as well ? possibly open to public
Artür Manukyan (08:32:45) (in thread): > or is this mostly internal ?
Estella Dong (08:33:50) (in thread): > Definitely public! And it’s a good idea to have period meeting
Artür Manukyan (08:35:09) (in thread): > I have seen this working pretty efficiently atscverse.zulipchat.com, and perhaps a similar approach can be taken here to establish standards
2024-09-10
Alex Qin (03:46:41): > @Alex Qin has joined the channel
2024-09-15
Jovana Maksimovic (23:19:51): > @Jovana Maksimovic has joined the channel
2024-09-23
Marcel Ramos Pérez (17:07:49) (in thread): > I just found out that youcanadd cross-organization issues by adding<org>/<repo>to the Search repositories field!@Dario Righelliand@Estella Dongshould add more issues to work on:pray:I’ve closed a couple already ^^
2024-10-07
Tuomas Borman (09:05:39): > @Tuomas Borman has joined the channel
shristi shrestha (11:07:18): > @shristi shrestha has joined the channel
2024-10-08
Alik Huseynov (03:10:05): > hybrid spatial hackathon - freehttps://spatialhackathon.github.io/interesting stuff on image-based ST data normalization and segmentation-free workflows - Attachment (ELIXIR-Germany SpaceHack): Home > Github Pages for SpatialHackathon
2024-10-09
Lambda Moses (10:23:50) (in thread): > It has satellite locations in Adelaide, Boston, and Memphis. I would consider going to Boston if more spatial people on the east coast are going.
Helena L. Crowell (11:01:54) (in thread): > Has anyone heard back from ESSB?:eyes:
Alik Huseynov (11:05:57) (in thread): > what exactly?
Helena L. Crowell (11:08:35) (in thread): > this conferencehttps://spatialbiologysociety.eu/is happening right after, also in Berlin - might be on purpose/co-organized, not sure… they said notifications would go out top of Oct., but haven’t heard back - Attachment (European Society for Spatial Biology): Homepage - European Society for Spatial Biology > Inaugural conference of the European Society for Spatial Biology (ESSB) Register for the conference 12 – 13 December 2024, Berlin, Germany Venue: Hotel & Conference Center Titanic Chaussee Berlin Welcome On December 12 and 13, 2024, the inaugural conference of the ESSB will take place here in Berlin. We expect an interdisciplinary audience from […]
Alik Huseynov (11:14:26) (in thread): > yes, the SpaceHack happening right before the ESSB, possibly on purpose. > I haven’t heard anything regarding the abstract/poster notifications either
Alik Huseynov (11:20:02) (in thread): > I can ask Denis Shapiro:wink:
Helena L. Crowell (11:20:27) (in thread): > If you do, say hi from me:wink:
Alik Huseynov (11:48:27) (in thread): > we will get notified in few days. Hi back from Denis
2024-10-11
Alik Huseynov (06:08:45) (in thread): > @Helena L. Crowellparticipating in SpaceHack 3.0?
Helena L. Crowell (06:11:42) (in thread): > hoping to … depends on ESSB a bit … things are a bit fiddly here wrt conferences/money etc.
Alik Huseynov (06:16:52) (in thread): > I see.. there is pre-ESSB event regarding FAIR spatial data -> on 11.12 between 15:00-19:00https://indico.dkfz.de/event/1112/ - Attachment (DKFZ Event Managment System (Indico)): 1st iO-FAST think tank meeting > WELCOME to the 1st IO-FAST think tank meeting- a satellite event of the ESSB conferenceIO-FAST (initiative for FAIR spatial data) aims to democratize spatial omics data to enable the community to share and (re)use spatial data in a FAIR manner. We will assemble leaders in the field and users to collect ideas, use-cases, pressing problems, solutions and most importantly provide a synergistic environment. With this community-driven effort we hope to introduce easy to apply frameworks and…
2024-10-16
Alik Huseynov (10:52:56) (in thread): > FYI, ESSB notifications on selected poster presentations are now out.
2024-10-22
Edward Zhao (03:08:57): > @Edward Zhao has joined the channel
2024-11-05
Chris (09:17:56): > @Chris has joined the channel
Chris (12:02:56): > Hi spatial community, > > Hope this question is still related in this channel. I try to annotate cell type from visium dataset of brain using a scRNA-seq reference. I apply the method used here:https://satijalab.org/seurat/articles/spatial_vignette. However, my result got zero score and I don’t know what went wrong. May I have your comment? Thank you! - Attachment (satijalab.org): Analysis, visualization, and integration of Visium HD spatial datasets with Seurat > Seurat - File (PNG): image.png
Jenny Drnevich (14:11:30) (in thread): > If you used the Allen reference subclass annotations as in the vignette, it means these two subtypes were predicted to have made up 0% of all spots. If you do not think that is possible, then you need to check your code for errors, see if the gene symbols don’t match up between the Allen reference and whatever the data set uses.
Chris (14:17:17) (in thread): > Hi@Jenny Drnevich, there were some warning messages but not error when I follow steps in the vignette. Will check again gene symbols in the reference and genes in the data set.
2024-11-06
Cindy Fang (09:55:49): > @Cindy Fang has joined the channel
2024-11-07
Dharmesh Dinesh Bhuva (17:57:20): > @Dharmesh Dinesh Bhuva has joined the channel
Malvika Kharbanda (22:11:29): > @Malvika Kharbanda has joined the channel
2024-11-08
Mike Smith (06:09:37): > @Mike Smith has joined the channel
2024-11-11
Chris (13:41:51) (in thread): > Hi@Jenny DrnevichWithAllen_reference@assays$RNAandcontrol@assays$Spatialare using gene symbol. Some cell types have signal but very weak. - File (PNG): spatialfeatureplot.png
Harry Mueller (18:08:10): > @Harry Mueller has joined the channel
2024-11-14
Davide Risso (05:22:23) (in thread): > Did you check the expression of some standard markers (such as Sst, Gad1, etc)? Did you look at what are the most highly expressed / spatially variable genes? Keep in mind that the Allen reference is heavily skewed towards neurons, so if you expect a large proportion of non-neuronal cell types it may not be the right reference.
2024-11-15
Chris (11:56:03) (in thread): > Hi@Davide Risso, thank you so much for your suggestion! When I plot neuron cell type such as L4 IT, L5 IT, I also didn’t see any signal which I expect must have. For SST, I got this. UsingSpatialFeaturePlot(), I didn’t find GAD1 or SLC17A6, SLC17A7. Here are cell types in the reference: > Pax6, L5/6 NP, L5 IT, L6 CT, L4 IT,Astrocyte, L2/3 IT, Vip, Sst Chodl, L6 IT Car3, L6 IT, Sncg, Pvalb,Oligodendrocyte, Lamp5 Lhx6, Sst, VLMC, Lamp5, Microglia-PVM, OPC, L6b,Endothelial, L5 ET, Chandelier, max. Cell types in bold have signal in my data. - File (PNG): image.png
Louise (15:00:53): > @Louise has joined the channel
2025-01-05
Alik Huseynov (09:05:18): > Does anybody know ifSpotCleanis compatible with Visium HD data or other similar tools exist to correct to diffusion?
2025-01-09
Ammar Sabir Cheema (11:41:00): > @Ammar Sabir Cheema has joined the channel
2025-01-10
Brian Repko (09:27:09): > @Brian Repko has joined the channel
2025-02-05
Wes W (11:35:39): > is there any software compatible with the new(ish) 2023 output from the Nanostring CosMx platform that is exported from Atmox? > > looks likeGiottoandvoyageronly support the old 2022 structures from my little play through , but if anyone has some dirty hacks or fixes to get it into angiottoobject like the old format that would be helpful > > i found Einan’s thread on github from 2023 and he wrote some really great scripts using ImageJ to reformat the tiffs into the right format, but my interpretation is he never really got the entire process to work with ceLLids > > I also read George’s reply’s to a 2024 thread on the same problem and issue. my understanding is that this is not yet resolved. although its something they would like to resolve. > > my previous experience in spatial has been merfish, Visim, and Xenium (for my bone marrows), first time doing a project with Nanostring’s 6000 plex CosMx platform, but generally, what are all you using to analyze and explore this data?
Lambda Moses (12:13:05): > OK, I’ll update my code to support the new version
Lambda Moses (12:22:24): > Are you talking about the basic CSV output or the TileDB stuff or the raw output that I see on the Nanostring website?
Wes W (12:24:36): > i cant speak to if the stuff on the nanostring site is recent. i know their “default” sample data they have that appears in most tutorials is the OLD 2022 format… but maybe they have added the new exported version… new output looks like this (multi screen shots incoming)
Wes W (12:24:39): - File (PNG): image.png
Wes W (12:24:58): - File (PNG): image.png
Wes W (12:25:23): - File (PNG): image.png
Wes W (12:26:06): > if thats the format on the site now, then great
Wes W (12:26:06): > inside those analysis FOVs looks like this - File (PNG): image.png
Wes W (12:26:08): > if you want my data
Wes W (12:26:12): > i’ll send it too you
Wes W (12:26:13): > haha
Wes W (12:26:51): > Cell comp folder contains jpgs
Lambda Moses (12:27:22): > The stuff on the website looks like more recent than the example dataset I initially used to develop thereadCosMXfunctionn in SFE, though I’m not sure if it’s the most recent. It would be great if you can share your data with me. How big is it? On the website the raw output is like over 300 GB.
Wes W (12:27:23): > csvs in Run summery
Wes W (12:28:14): > let me see
Wes W (12:30:18): > okay well if i use ztfs compression i could get it below 300GB
Wes W (12:30:42): > for one sample
Wes W (12:31:42): > sending you DMs
Lambda Moses (12:33:09): > So it looks like the raw outputs.readCosMXis for the basic CSV outputs. Do you still get the CSV files from AtoMX?
Wes W (13:03:02): > AtoMX exports this data format that contains many csvs and summary files as well as the new ‘raw’ files
Wes W (13:03:12): > but the formats are different
Wes W (13:03:16): > including the tiffs and jps
Wes W (13:03:20): > from old for images
2025-02-06
Davide Risso (08:39:34): > Q: do you absolutely need the images in your R object? Because, we found that for most of our analyses the count matrix and the polygon file are enough. Sometimes we use the transcript file but not always.
Helena L. Crowell (08:45:46): > Second that.The flat file export module we use in AtoMx gives a simple set of files per slide - counts, metadata, polygons, transcripts, fov positions - that are straightforward to read into anything you like.These data are also considerably lighter (couple GB, not 100s of GB), and sufficient unless you are working on methods devel for spot calling or segmentation.
Kalyanee (09:21:05): > @Kalyanee has joined the channel
Wes W (17:17:52): > or like cell morphology
2025-02-10
Davide Risso (09:03:20): > Polygons should be enough for cell morphology? Or am I misunderstanding your comment?
Jayaram Kancherla (18:39:25): > Hi all, I was wondering if anyone here tried to convert spatial data to SFE/SPE and vice-versa?
2025-02-11
Alik Huseynov (01:18:09): > SPE to SFE is pretty straightforward
Lambda Moses (01:23:55) (in thread): > I think Jayaram is talking about converting between SpatialData and SFE
Alik Huseynov (01:27:00) (in thread): > That’sone is not implemented yet
Jayaram Kancherla (01:44:56) (in thread): > yes spatial data to SFE.
Alik Huseynov (06:40:55) (in thread): > I’m not awareifit is already available, see this discussion ->https://community-bioc.slack.com/archives/C0558SHL814/p1736012170685549?thread_ts=1727262555.519629&cid=C0558SHL814 - Attachment: Attachment > Hi both @Luca Marconato and @Helena L. Crowell. Any updates on reading SpatialData object in R? > Eg, I would like to load existing xenium_test.zarr object, then make SpatialExperiment or ideally SpatialFeatureExperiment object for downstream spatial analysis. > If there is some preliminary workflow/vignette to play around, please do share. Thanks!
Artür Manukyan (13:59:30) (in thread): > Yeap, still a work in progress.
Wes W (14:27:26) (in thread): > you are. but thats okay. one step at a time. still struggling to getreadCosMX()to read the new flat files from CosMx’s Atomx output
Kevin (15:20:35): > @Kevin has joined the channel
Wes W (15:44:24): > okay so, am I correct in assuming everyone else has figured out a hack solution to read in the CosMx flat files into Giotto and that its so easy that I am missing something obvious that no one has posted a solution?
Helena L. Crowell (16:00:28) (in thread): > Maybe im missing sth, but the Table is already a SCE, and points/shapes are linked to .parquet, so assuming SFE just needs those paths for the molecules/boundaries, wouldnt that be pretty easy already with what’s there? Unless SFE needs the data in a specific (different) format. Then I dont see a way of doing a comversion without duplicating on disk?Or realizing everything…
Helena L. Crowell (16:12:37) (in thread): > No, it’s not that easy.1- I think (at least to me) it’s not clear exactly what data you need.Perhaps you could help by clearly explaining? Is it the images from IF stains? I know CosMx provides a composite, but to my knowledge, individual channels (eg DAPI, CD45…) would only be available after running their napari plugin? > 2- As you said that polygons are insufficient, perhaps the reason for lack of a “solution” is because most image analysis currently lives in Python. So handling the images is maybe missing/not well addressed by R/Bioc infrastructure.But clarifying (1) would really help with understanding what needs to be done better:pray:
Wes W (17:21:51) (in thread): > number 2 I will solve later. as i want the images for cell morphology as the way we tell the difference between the cancer cells and not cancer cells is their shape and writing scripts to measure cell expression based on distance to a morphology or changes in amounts of morphologies can come later…. > > getting back to number 1, ignoring the raw data, and just using the flat files from CosMx (exprMat, fov_positions, metadata, tx_file, and polygons) > > manually creating an object fromcreateGiottoObject()its no immediately clear the formats to make giotto happy for downstream > > obviously some stuff is pretty straight forward from just my experience using Giotto for my Xenium data… i need to convert the CosMx expression matrix rows to coloumns as Giotto expects cell IDs as column names and CosMx puts gene names as coloumns… thats easy… > > but figuring out which fovs to toss based on global mm or global px isn’t as easy, as giottoObject just wants one of those, but i have a feeling to get my images to line up later I am going to need a specific one of those… > > like I am honestly just struggling to coerce the CosMx output into an Giotto object that can then be used on the Giotto pipeline at the point… > > i’m going to take a dinner break and give it another go. i feel like i am close, but i also felt like i was close like 20 times already haha
Artür Manukyan (17:25:27) (in thread): > @Ruben Dries
Ruben Dries (17:44:27) (in thread): > @Jiaji George Chenis the best positioned to help you with this issue
Ruben Dries (17:51:23) (in thread): > @Wes Wyou probably already looked at those but just in case: > * example with cosmx datahttps://drieslab.github.io/Giotto_website/articles/nanostring_cosmx_lung_cancer.htmlcosmx has multiple output formats, so it’s not guaranteed to work > * general information about creating a giotto object:https://drieslab.github.io/Giotto_website/articles/object_creation.htmlthis should in theory work for any type of spatial data (including subcellular like cosmx,https://drieslab.github.io/Giotto_website/articles/obj_create_pnts.html)
Ruben Dries (17:51:52) (in thread): > Also looping in@Joselyn Chávezwho might also be able to help
Wes W (19:49:39) (in thread): > @Ruben Driesthanks for those links. I had seen all but one of them, I will have a read on it > > the cosmx lung cancer one is sadly no longer vaild. thats the old cosmx data output format from that went away in 2022. the 2023 changed it and the updated 2024/2025 output for CosMx from AtoMx doesnt even use the same id orinination in the image data anymore its kinda a whole thing > > see:https://github.com/drieslab/Giotto/discussions/783and:https://github.com/drieslab/Giotto/issues/785as unresolved similar issues =( > > but flat files exist now so hopefully I can solve this year long issue - Attachment: #783 Nanostring CosMx and AtoMx exported data - Attachment: #785 Nanostring CosMx and AtoMx exported data
Wes W (20:03:31) (in thread): > happy to upload the outputs from modern CosMx if anyone wants a play around
Jiaji George Chen (20:10:00) (in thread): > Hi@Wes W, the currentimportCosmx()utility (https://drieslab.github.io/Giotto_website/articles/import_utilities.html) andcreateGiottoCosMxObject()only work for the original outputs, like you mentioned. That would be the legacy lung dataset, the WTX human lymph example they have up, and the human lymph dataset, with a little wrangling of the polygon csv they provide. > > Please do upload! I’ll take a look and update it.
Wes W (20:11:15) (in thread): > easy! whats your email, i’ll send you a box share link
Wes W (20:13:26) (in thread): > sent
Wes W (20:13:33) (in thread): > you should have access to the flat files now
Jiaji George Chen (20:15:01) (in thread): > Got it
2025-02-12
Alik Huseynov (00:43:50) (in thread): > Need to give it a try
Jiaji George Chen (12:02:27) (in thread): > To update here: > * importCosMx()is currently working for the flat file outputs for all but the polygon csv. > > * if a folder with all the CellLabels is manually assembled from the raw outputs and placed with the other flat files similarly to the old outputs (same with CellComposite)createGiottoCosMxObject()might already work out of the box > > * Polygons are pretty simple to load separately from adata.framewithcreateGiottoPolygon(), but I’ll update to better support this. > * An update to the way thatimportCosMx()$load_images()grabs filepaths should be enough to map the images from the raw output
2025-02-13
Jayaram Kancherla (00:26:03) (in thread): > Hi Helena, would be great if you can help pin the version of python packages to load the test dataset from the package:https://github.com/HelenaLC/SpatialData/issues/112second, I think it would be great to coordinate and agree on a set of rules to coerce between these objects, uniformly across different packages and efforts.
Helena L. Crowell (02:06:56) (in thread): > Could you mention that in a GH issue? Slack is not a good place for me to track software issues.Also so wee keep a record, others can join discussion etc. Thanks!:heart_hands:
Helena L. Crowell (04:22:59) (in thread): > I’m actually not sure by what you mean by “rules”.Can you explain? Wouldn’t the conversion be dependent on the target class?
Artür Manukyan (05:30:36) (in thread): > @Wes WI think examples of new flat files are not publicly available anywhere by Nanostring right ?
Wes W (11:40:49) (in thread): > you are correct, if you want to play with my files@Artür Manukyan, i can send them too you. > > we are like halfway there forspatialfeatureexperimentand forgiottogene expression and polygons are working. getting the object to build is still a little struggle =(
2025-02-21
Gabriel Grajeda (09:34:33): > @Gabriel Grajeda has joined the channel
Gabriel Grajeda (10:22:24): > Hi everyone, I’m working on adding RCTD (a cell type deconvolution method) to Bioconductor. RCTD uses an annotated RNA-seq reference dataset to deconvolve spatial transcriptomics data. The original package, which I did not author, allows users to optionally specify their own UMI counts rather than using the column sums of the count matrices. > > I was wondering if this feature is common practice, and whether there is a standard approach for supporting custom UMI counts through Bioconductor classes likeSummarizedExperimentandSpatialExperiment. My initial thought is to allow users to specify their UMI values in thecolData, but I wanted to see if there is already a convention for this. > > The modified RCTD package can be foundhereif anyone is interested. I would appreciate any help on this!
2025-02-24
Peter Hickey (16:23:18): > > allows users to optionally specify their own UMI counts rather than using the column sums of the count matrices > I don’t really understand your question. > The column sums of a count matrix are often referred to as the ‘library sizes’ - is that what you mean? > UMI counts are themselves a count matrix
2025-02-26
Vince Carey (06:04:29): > I have a question about data curation in the spatial domain. IMHO scRNAseq provides valuable patterns for conveying curated experiments to users. Inspired by@Lambda Moses’ alabaster.sfe, I am asking whether aspects of the newer scRNAseq framework should be adopted for conveying spatial omics experiments. User-facing components of the scRNAseq framework include surveyDatasets(), which interrogates an ArtifactDb metadatabase for metadata regarding provenance, curation date, assay types, volume, listDatasets(), which is more compact, and per-experiment functions that acquire and cache serialized experiments. With the ArtifactDb discipline, data creators have opportunities to improve accessibility and FAIRness of their work. But we need to make a concerted effort to understand and use the underlying schemas, validators, and transfer APIs. Note that this approach gives python programmers comparable access to data and metadata, with dolomite replacing alabaster. Check out the#artifactdbchannel for related information.
Vince Carey (06:08:25): > I am tagging@Jayaram Kancherla@Aaron Lunfor additional input. I am aware of initiatives called IOFAST and GESTALT that build communities for methods in spatial omics but I do not have bandwidth to follow them and to understand how they are looking at curation and metadata. I do feel that the example data that are readily available are somewhat lacking in metadata about experimental context.
Artür Manukyan (07:16:29) (in thread): > For those who might be interested: > IO-FAST:iofast.zulipchat.comis the zulip channel for discussions > GESTALT:https://globalspatial.org/join-gestalt/main website
2025-02-27
Jayaram Kancherla (10:46:22) (in thread): > I would love to get involved once i am back from my leave (beginning of april). I believe as more and more of these datasets become accessible, efforts like cellxgene or single-cell portal for spatial will probably become more prevalent. I also think before things get there, formalizing the on-disk formats in R and Python for the same dataset is crucial. In the cellxgene case, the h5ad’s and RDS files contain slightly different information and naming structures.
2025-02-28
Vince Carey (05:08:50): > And this comprehensive overview of SFE also mentions alabaster.sfe… nice work@Lambda Moses! - Attachment (bioRxiv): Geospatially informed representation of spatial genomics data with SpatialFeatureExperiment > SpatialFeatureExperiment is a Bioconductor package that leverages the versatility of Simple Features for spatial data analysis and SpatialExperiment for single-cell -omics to provide an expansive and convenient S4 class for working with spatial -omics data. SpatialFeatureExperiment can be used to store and analyze a variety of spatial -omics data types, including data from the Visium, Xenium, MERFISH, SeqFish, and Slide-seq platforms, bringing spatial operations to the SingleCellExperiment ecosystem. > > ### Competing Interest Statement > > The authors have declared no competing interest.
2025-03-07
Emanuele Pitino (05:36:45): > @Emanuele Pitino has joined the channel
2025-03-11
Andrew Ghazi (11:06:38): > @Andrew Ghazi has joined the channel
2025-03-13
Mihai Todor (06:59:40): > @Mihai Todor has joined the channel
2025-03-18
Julien Roux (05:09:10): > @Julien Roux has joined the channel
Nicolo (15:01:27): > @Nicolo has joined the channel
2025-04-07
Carlo Pecoraro (08:43:28): > Upcoming online course on Spatial Omics with R/Bioconductor, taking place in Mayhttps://www.physalia-courses.org/courses-workshops/spatial-omics-1/ - Attachment (physalia-courses): spatial Omics in R/Bioconductor > 19-21 May 2025 To foster international participation, this course will be held online
Jenny Drnevich (11:26:28): > Anyone working on Z-stack/3D reconstruction of spatial images? Any recommendations to existing packages/workflows would be greatly appreciated!
Artür Manukyan (11:28:47) (in thread): > You can check our work if this suits youhttps://bioinformatics.mdc-berlin.de/VoltRon/registration.html
Artür Manukyan (11:28:56) (in thread): > uses R-native operations, would appreciate feedback too
Jenny Drnevich (11:35:51) (in thread): > Looks perfect - even does Xenium!! I’ll give feedback on your GitHub. Thanks so much!
Lambda Moses (19:04:07): > When I use thespatialDEpackage, it fails to install the SpatialDE package Python package. The culprit is that scipy can’t be installed due to dependency conflicts:https://github.com/sales-lab/spatialDE/blob/75f178cd26471c448e5f35a21ac048d9eb821c94/R/basilisk.R#L3This is the error I got when basilisk is trying to install:package scipy-1.9.3-py38h8ce737c_0 requires python_abi 3.8.* *_cp38, but none of the providers can be installed > > Could not solve for environment specs > The following packages are incompatible > ├─ python 3.12.8** is installable and it requires > │ └─ python_abi 3.12.* *_cp312, which can be installed; > └─ scipy 1.9.3** is not installable because there are no viable options > ├─ scipy 1.9.3 would require > │ └─ python_abi 3.8.* *_cp38, which conflicts with any installable versions previously reported; > ├─ scipy 1.9.3 would require > │ └─ python_abi 3.10.* *_cp310, which conflicts with any installable versions previously reported; > ├─ scipy 1.9.3 would require > │ └─ python_abi 3.11.* *_cp311, which conflicts with any installable versions previously reported; > └─ scipy 1.9.3 would require > └─ python_abi 3.9.* *_cp39, which conflicts with any installable versions previously reported. >
2025-04-08
Davide Risso (06:06:04) (in thread): > @Gabriele Sales@Dario Righelli
Gabriele Sales (06:07:19): > @Gabriele Sales has joined the channel