pirl-unc / hitlist / 27080187832
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 29
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: 80.39% (+0.02%) from 80.371%
27080187832

push

github

web-flow 
v1.31.3: classify EBV-LCL mono-allelic hosts (721.221/C1R) as ebv_lcl, not cancer (#280)

Surfaced while reviewing the #33b curation: ~313K observations of
mono-allelic immunopeptidome data on EBV-LCL transfection hosts were
mislabeled src_cancer=True — including Sarkizova 2020 (277K rows),
Abelin 2017 (27K), Hilton 2017 (9K). And it was INCONSISTENT: C1R
studies (Marcilla, Illing) were correctly ebv_lcl while 721.221 studies
were 'cancer', purely because IEDB tags some hosts' culture condition
'EBV transformed, B-LCL' and others plain 'Cell Line / Clone'.

Root cause: classify_ms_row derived is_ebv_lcl only from IEDB's
(inconsistent) culture_condition. 721.221 and C1R are EBV-transformed
B-LCLs (Cellosaurus CVCL_6263 / CVCL_3714, 'Transformed cell line' from
normal blood) — immortalized B cells used as single-allele transfection
hosts, not tumors.

Systemic fix (not per-PMID): an 'ebv_lcl: true' flag on the host entries
in monoallelic_lines.yaml + classify_ms_row treating mono-allelic data on
a flagged host as ebv_lcl (src_cancer=False), independent of IEDB's
tagging. One change auto-corrects all 721.221/C1R studies, existing and
future. Genuinely malignant hosts (K562 = CML leukemia) are not flagged
and keep src_cancer=True.

Verified: Sarkizova/Abelin/Hilton/Schittenhelm -> ebv_lcl/not-cancer;
K562 -> stays cancer; HeLa unaffected. Full suite 873 passed.

5071 of 6308 relevant lines covered (80.39%)

0.8 hits per line

Coverage Regressions

Lines Coverage ∆ File
17
91.79
 0.09% curation.py
Jobs
ID Job ID Ran Files Coverage
1 27080187832.1 29
80.39
 GitHub Action Run
Source Files on build 27080187832