pirl-unc / hitlist / 24615976599
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 19
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: 43.981% (+0.07%) from 43.91%
24615976599

push

github

web-flow 
v1.10.4: Fix Liepe 2016 fibroblasts + Caron 2015 PBMCs escaping cell_line override (#71) (#78)

Closes #71. The cell_line override path already correctly handles
EBV-LCLs (is_cancer = not is_ebv_lcl) for rows IEDB flags as
"Cell Line / Clone (EBV transformed, B-LCL)". So the 17-PMID "mass
misclassification" described in issue #19 is largely obsolete — most
affected studies are already correctly classified because IEDB does
flag them.

The real remaining problem is narrower: PMIDs that mix cell-line samples
with Direct Ex Vivo primary cells in IEDB. The PMID-level
`override: cell_line` fires for ALL rows regardless of culture_condition,
forcing is_cancer=True on the primary cell rows.

Audit of override:cell_line PMIDs with Direct Ex Vivo rows:
  PMID 28514659: 9,292 splenocyte rows (possibly mouse)
  PMID 31844290: 6,981 CLL-patient PBMC rows (actual cancer — correct)
  PMID 26154972: 4,617 primary PBMC rows (no disease — wrongly cancer)
  PMID 27846572: 3,614 primary fibroblast rows (no disease — wrongly cancer)
  PMID 30779550: 3,001 splenocyte rows (possibly mouse)
  PMID 38480730: 1,168 rows marked disease=healthy
  PMID 27600516: 79 melanoma-patient PBMC rows (cancer — correct)

This PR fixes the two clearest cases — primary human fibroblasts and PBMCs
with no disease label — via per-sample rules. Splenocyte / mouse / healthy
cases are left for a follow-up once their paper context is verified.

Changes:
- PMID 27846572 Liepe 2016: added rule routing Direct Ex Vivo rows through
  `override: healthy` (primary fibroblasts are non-malignant comparators,
  not cancer).
- PMID 26154972 Caron 2015: added same rule for Direct Ex Vivo PBMCs
  (primary PBMC from healthy donors) + a matching ms_samples entry.

Regression tests added:
- test_liepe_2016_fibroblast_not_cancer
- test_caron_2015_pbmc_not_cancer

1235 of 2808 relevant lines covered (43.98%)

0.44 hits per line

Jobs
ID Job ID Ran Files Coverage
1 24615976599.1 19
43.98
 GitHub Action Run
Source Files on build 24615976599