25175043390
80%

Ran 30 Apr 2026 03:48PM UTC

Jobs 1

Files 26

Run time 1min

Badge

Embed ▾

Committed 30 Apr 2026 03:45PM UTC coverage: 70.482% (+0.04%) from 70.447%

Build # 25175043390

Build Type

push

github

Committed by

web-flow

Commit Message

v1.30.14: Tokenize multi-allele genotype strings in _flatten_hla_alleles (#199)

~32% of ms_samples in pmid_overrides.yaml encode a donor's full
genotype as one space-separated string (e.g. "HLA-A*01:01 HLA-B*07:02
HLA-C*12:03"). _flatten_hla_alleles was treating those strings as
single phantom "alleles" because they passed the syntactic check
(start with HLA-, contain * and :).

Effect on the new qc reports:
- cross_reference reported the entire genotype string as a yaml_only
  divergence — confused curators looking for typos.
- curation_plan inflated yaml_only_alleles_n by 32%, skewing the
  priority queue. Bassani-Sternberg 2017 shows 42 → 32 yaml-only
  alleles after fix; Marcu 2021 (which had its real signal masked
  by the noise) drops out of the top-10 entirely.

Fix: split the input on whitespace and keep only tokens that pass
the per-token allele check. Drops noise tokens like 'or' / commas
in free-text fields silently rather than mistaking them for alleles.

Also silences a pandas 3.0 FutureWarning in curation_plan's fillna
path by going through pd.to_numeric.

Three new tests in test_curation.py: multi-allele split, single-
allele non-regression, and noise-token filtering.

Coverage Stats

3584 of 5085 relevant lines covered (70.48%)

0.7 hits per line

Coverage Regressions

Lines	Coverage	∆	File
10	88.28	0.18%	curation.py
1	93.06	0.0%	qc.py

Jobs

ID	Job ID	Ran	Files	Coverage
1	25175043390.1	30 Apr 2026 03:48PM UTC	26	70.48	GitHub Action Run

pirl-unc / hitlist / 25175043390
80%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Coverage Regressions

Jobs

Source Files on build 25175043390

pirl-unc / hitlist / 25175043390 80%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Coverage Regressions

Jobs

Source Files on build 25175043390

pirl-unc / hitlist / 25175043390
80%

README BADGES
x