Mitchell_T1D: individual ancestry replaces the conversion-time constant, and age is filled in for 53 specimens. Emerson_CMV: ancestry corrected for 15 specimens. Both cohorts gain source race/ethnicity and provenance columns. The immuneML metadata and the mitchell-t1d, t1d-pooled and cmv task files, MANIFEST, DATASHEET, dataset card and immuneML README are updated to match. Labels, folds and repertoires are unchanged.
8 days ago
Remove the exact specimen date from Rawat_T1D metadata
The Rawat_T1D metadata carried each specimen's exact collection date for all 614 participants. No benchmark task reads it, and combined with the day-precise age column it reveals each participant's date of birth. The column is removed; all 614 rows and the other 36 columns are unchanged. DATASHEET.md records the removal. Earlier commits still contain the values.
Describe Mal-ID rows as individual reads, not deduplicated clones
Mal-ID repertoire files have one row per sequencing read (identical reads repeat; clone_id groups reads into clones), while immunoSEQ files have one row per unique rearrangement with its count. Updates DATASHEET.md and the card's file description accordingly.
Add immuneML-ready AIRR export (immuneML/) and document it in the cards
Adds immuneML/: AIRR Rearrangement repertoires for all 2,352 specimens, immuneML metadata CSVs, fixed fold splits and example specs for the 11 benchmark tasks, MANIFEST.tsv and export_report.json (built from 165c551 with utils/immuneml_export.py and utils/immuneml_tasks.py). README.md, DATASHEET.md and croissant.json describe the export.
9 days ago
Remove remaining 'Anonymized during review' placeholders from datasheet
