Curator Desk
BioAnalyzer curator desk
Use this page to review BioAnalyzer’s predicted statuses for candidate papers and record curator ground truth for each PMID.
What you can do here:
- Find papers quickly: Search, sort, and filter the table (Priority score helps surface high‑value candidates).
- Record ground truth: Enter curator TRUE labels and optional “was the prediction correct?” flags per field.
- Share feedback: Submit feedback via a GitHub issue (CSV is pre-filled in the issue body).
The table is built from data stored in the data/ folder. To analyze your own papers:
- Place your CSV (or Parquet) file in
data/(e.g.,data/sample.csv). - Run
quarto render.
The system reads data/sample.csv by default. To use a different source, set the CURATOR_DATA_PATH environment variable to your specific path (e.g., data/analyzed_papers.csv) or use CURATOR_DATA_URL for remote files.
Expected columns
- Required:
PMID - Recommended:
Title,Journal,Year,Summary - Status values:
ABSENT|PARTIALLY_PRESENT|PRESENT
Candidate Curatable Articles
Tip: Sort by Priority (descending) to review the most promising candidates first. Use the search box and column filters to narrow your focus.
Use the table above to find papers (the Open button in the Published column opens the paper on PubMed). Then select or enter a PMID below; the page will scroll to the field-by-field validation form where you can enter your review. Feedback uses the same schema as the Streamlit app; because this site is static, saving is done by downloading a CSV per row—submit via your usual process (e.g. email or GitHub issue).
- Quick select a PMID from the dropdown, or type a PMID and click Load row (the form will scroll into view).
- For each field, review BioAnalyzer predicted and set Curator TRUE label and Was BioAnalyzer correct?
- Click Download feedback CSV to save one row; submit as needed.
Curator feedback form
Curator ID / initials
Quick select for feedback — Select a PMID from the table, or type below.
Title (from table)
Overall paper verdict
BioAnalyzer version (fixed for all submissions)
Field-by-field validation (ground truth)
For each field, provide the curator TRUE label (ground truth). Optionally mark whether BioAnalyzer’s predicted status was correct.
Select a PMID above and click Load row to see BioAnalyzer’s prediction.
Curator TRUE label for Host Species: Was BioAnalyzer correct for Host Species?
Select a PMID above and click Load row to see BioAnalyzer’s prediction.
Curator TRUE label for Body Site: Was BioAnalyzer correct for Body Site?
Select a PMID above and click Load row to see BioAnalyzer’s prediction.
Curator TRUE label for Condition: Was BioAnalyzer correct for Condition?
Select a PMID above and click Load row to see BioAnalyzer’s prediction.
Curator TRUE label for Sequencing Type: Was BioAnalyzer correct for Sequencing Type?
Select a PMID above and click Load row to see BioAnalyzer’s prediction.
Curator TRUE label for Sample Size: Was BioAnalyzer correct for Sample Size?
Submit review — Opens GitHub to create an issue with your feedback (no file download). A bot will then open a Pull Request with the feedback CSV so the maintainer can view and download it (including your initials and comments).
Notes: Submissions are processed as issues and a tab will be opened for you to just click the create button. Reviewers will be looking into you submittd issue to evaluate the results