pirl-unc / hitlist / 24913437735
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 23
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 58.903% (+0.04%) from 58.862%
24913437735

push

github

web-flow 
v1.18.1: Gomez-Zepeda per-file dedupe + Raji/plasma curation (#147) (#154)

Issue #147: supplementary rows were deduped on
(peptide, mhc_restriction, pmid), collapsing peptides that are
genuinely presented by multiple cell lines in the Gomez-Zepeda 2024
5-sample panel onto one arbitrary sample context.  Local audit showed
the current dedupe would drop 5,109 Gomez rows (2,982 with non-empty
mhc_restrictions).

- hitlist/supplement.py:
  - Add a supplementary_file column to every emitted row so
    per-sample provenance is preserved downstream.
  - Fold the filename into the synthesized assay_iri
    (supplement:PMID:FILE:peptide:mhc) — still row-unique after
    the dedupe change, still works with issue-#146's evidence_row_id
    contract.
  - Dedupe key now includes supplementary_file, so within-file
    duplicates still collapse but the JY vs Raji vs HeLa vs SK-MEL-37
    vs plasma rows for the same peptide survive distinctly.

- hitlist/data/pmid_overrides.yaml (PMID 38480730):
  - Add a 'Raji (Burkitt's lymphoma EBV-LCL)' unperturbed sample
    carrying the curated HLA genotype
    (A*03:01 / B*15:10 / C*03:04 / C*04:01) so the 17,328 Raji
    supplementary rows with predicted allele restrictions can join
    a real sample instead of falling through to the PMID class pool.
  - Upgrade the existing Raji-spike sample from 'HLA class I' to the
    same exact alleles (spike transfectants share host genotype).
  - Add a 'Human plasma (commercial pool)' class-only sample so the
    973 plasma predicted-allele rows attach sample context.  Donor
    HLA typing is not reported, so the sample stays class-only; per-
    row restrictions still provide allele evidence where present.
  - HeLa HLA-A*03:19 rows are intentionally left falling through
    to the PMID class pool pending a supplement re-audit (256 rows;
    HeLa curated typing is A*68:02 / B*15:03 / C*12:03 per the paper's
    Methods).

Tests (6 new in tests/test_supplement.py):
- supplementary_file ... (continued)

2405 of 4083 relevant lines covered (58.9%)

0.59 hits per line

Coverage Regressions

Lines Coverage ∆ File
2
87.5
 0.74% supplement.py
Jobs
ID Job ID Ran Files Coverage
1 24913437735.1 23
58.9
 GitHub Action Run
Source Files on build 24913437735