pirl-unc / hitlist / 25189785861
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 27
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: 70.9%. Remained the same
25189785861

push

github

web-flow 
v1.30.20: Strip PTM annotations from peptide before length-based severity tiering (#206)

Pre-v1.30.10 parquets carried IEDB's inline PTM annotations directly
in the peptide column ("CGPSGLVREL + METH(C1)" instead of the bare
"CGPSGLVREL"). The v1.30.17 severity classifier was measuring the
full annotated string length, so a 10-aa class-I peptide modified
with methylation became a 21-character "implausible" row.

Effect on the production parquet: ~36k Sarkizova 2020 (PMID 31844290)
rows were tagged implausible purely on PTM-suffix length, falsely
pushing that study to the top of the curation_plan triage queue.
Faridi 2020, Goncalves 2021, Mei 2020, Mifsud 2021 and others
exhibited the same pattern.

Fix: split on " + " (regex=False -- the literal "+" is otherwise
parsed as a regex quantifier and the split silently no-ops) and
take the prefix length. v1.30.10+ parquets where peptide is already
bare are unaffected -- the split returns the original string. Adds
a regression test exercising METH, DEAM, no-PTM, genuinely-long, and
malformed-PTM paths.

Drive-by: fix test_load_pmid_overrides_warns_on_legacy_keys monkey-
patch to delegate non-pmid_overrides.yaml lookups to the real
_data_path. Previously the monkeypatch returned a bare filename for
monoallelic_lines.yaml, which load_pmid_overrides now validates
against -- the test only passed when the lru_cache was warmed by an
earlier test (order-dependent flake under pytest -n auto).

3662 of 5165 relevant lines covered (70.9%)

0.71 hits per line

Coverage Regressions

Lines Coverage ∆ File
1
94.52
 0.0% observations.py
Jobs
ID Job ID Ran Files Coverage
1 25189785861.1 27
70.9
 GitHub Action Run
Source Files on build 25189785861