pirl-unc / hitlist / 33803078776
84%

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: 82.927% (+0.4%) from 82.522%
33803078776

push

github

web-flow 
v1.56.0: truthful sample-MHC attribution — centralize the parse, split pooled genotypes (#417)

* Centralize sample-MHC attribution and split pooled genotypes

Adds a single API for turning a curated ms_samples[].mhc field into
attribution candidates, and fixes the curated samples whose mhc field
pooled several donors into one genotype.

curation.sample_mhc_candidates / SampleMhcCandidates keep the three
precisions apart: exact molecules, serotype designations plus their
expanded members, and locus/class designations that name no allele.
Before this, everything but the first yielded the empty set, so seven
samples across five PMIDs dropped out of the allele-level join silently
(#380).  The export join uses join_alleles so a serotype-typed sample
matches its members, marked sample_attribution="serotype_expansion"
rather than "allele_exact"; the peptide summary keeps reporting a
serotype as a serotype.

sample_alleles_for_pmid is now public — it is the direct answer to "what
did this study type its samples to?", and was previously reachable only
through the per-peptide functions, which are empty for most studies.

qc.sample_ploidy_audit flags any curated sample carrying more than two
alleles at one locus.  A diploid donor cannot, so it is proof the mhc
field pools several samples.  It found six such samples; each is fixed
here from the primary source:

- 36423003 (#381): split into 8 per-cell-line samples.  The 13 class-I
  alleles are a union across eight animals of six BoLA haplotypes.  Note
  records that peptide-to-allele assignment is NetMHCpan-predicted, not
  measured, and that class II stays locus-level because no line-to-DRB3
  mapping exists.
- 32350084: split into 19 EBV-LCL + 7 K562 transfectant samples.  The
  previous 6-allele list came from the paper's abstract (a clustering
  result), not its Methods; three of those alleles were never transduced
  into K562.  DPA1 is now curated with each beta chain.
- 26768311: split into 10 mono-alleli... (continued)

6708 of 8089 relevant lines covered (82.93%)

0.83 hits per line

Coverage Regressions

Lines Coverage ∆ File
128
89.08
 1.25% export.py
33
94.4
 0.44% curation.py
15
94.14
 0.53% qc.py
Jobs
ID Job ID Ran Files Coverage
1 33803078776.1 30
82.93
 GitHub Action Run
Source Files on build 33803078776