pirl-unc / hitlist / 33231629012
81%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 30
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: 81.034% (+0.2%) from 80.865%
33231629012

push

github

web-flow 
v1.49.0: make curated sample metadata self-consistent (#372, #374, #375, #379) (#378)

* v1.49.0: make curated sample metadata self-consistent (#372, #374, #375)

Three defects found by cross-checking every curated ms_samples entry
against the class/species mhcgnomes derives from its own alleles.

#372: species was resolved once per study, outside the sample loop, so
sample.get("species") was never read. Both samples carrying the key are
mouse arms of otherwise-human studies with no study-level species, so
they defaulted to and exported as Homo sapiens. Now resolved per sample
with the study value as fallback.

#374 (part): 721.221-HLA-G*01:01/03/04 were declared classical "I".
HLA-G is class Ib; the corpus already curates the sibling HLA-E/HLA-F
and C1R HLA-G samples as non-classical, so this was an internal
inconsistency that also made --mhc-class I return them.

#375: three curated mhc values contained tokens mhcgnomes cannot parse
and therefore yielded no alleles at all — BL2*02 (the chicken class-II
B locus is BLB, so BLB2*02), and SLA-I / BoLA-I, which are class
designations written in gene position (SLA class I / BoLA class I).

Guard tests pin all three: every non-class-only mhc must yield at least
one allele, no declared class may contradict its alleles, and the
per-sample species override must be honored. The 12 remaining class
contradictions are allow-listed with their issue reference rather than
hidden — Patr-AL is genuinely unsettled and the 11 "I+II" samples need
their class-II genotypes read out of each paper.

Closes #372, closes #375

Claude-Session: https://claude.ai/code/session_014cBpazLoFvftLSLavT7T9X

* Correct species-inference traps; validate species per sample (#379)

Follow-up to the review of the #375 fixes. Three of the "obvious"
corrected tokens silently resolved to the wrong species, because
mhcgnomes infers a species for unprefixed or generic input:

  BLB2*02        -> Coja-BLB2*02 (Japanese quail), not chicken
  BoLA c... (continued)

6097 of 7524 relevant lines covered (81.03%)

0.81 hits per line

Coverage Regressions

Lines Coverage ∆ File
141
86.22
 0.56% export.py
122
51.78
 0.0% cli.py
45
92.65
 0.5% curation.py
9
96.11
 0.02% observations.py
Jobs
ID Job ID Ran Files Coverage
1 33231629012.1 30
81.03
 GitHub Action Run
Source Files on build 33231629012