pirl-unc / hitlist / 25389171834
75%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 28
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: 71.522% (+0.09%) from 71.434%
25389171834

push

github

web-flow 
v1.30.37: is_non_peptide_ligand — exclude CD1/MR1/MIC rows from peptide loaders (#228) (#229)

* v1.30.37: is_non_peptide_ligand — exclude CD1/MR1/MIC rows from peptide loaders (#228)

CD1, MR1, MIC{A,B}, RAET1*, ULBP*, NKG2[A-C], and HFE are not
peptide presenters — they bind lipids, glycolipids, or
small-molecule metabolites.  IEDB curates these alongside
peptide-MHC and stores chemical names / compound identifiers
in the ``peptide`` column.  In v1.30.36 the corpus carried 322
such rows that all passed existing severity / suspect filters
and silently flowed into peptide-prediction models, motif
analyses, and length distributions.

Add ``is_non_peptide_ligand`` (computed at scan time over
``mhc_restriction``) on observations + binding + training
exports.  Default ``load_observations`` /
``generate_observations_table`` / ``generate_binding_table``
filter these rows out; opt back in with
``exclude_non_peptide_ligand=False``.  H2-M3 is intentionally
excluded from the regex — it presents N-formyl peptides.

Validated: regex matches the documented 322 rows on the v1.30.36
corpus with no classical / class-Ib false positives.

* Annotate is_non_peptide_ligand intent: defensive NKG2, intentional triple derive

Doc-only follow-up from PR review.

- curation.py: NKG2 is a NK-receptor family, not an MHC molecule, and
  should never appear in IEDB's MHC-restriction column. Kept in the
  regex defensively (matches the issue spec) but explicitly labeled.
- observations.py + export.py: spell out that is_non_peptide_ligand is
  derived in three places by design, not by accident — scan time, load
  time (here), and _apply_training_defaults — so a future reader doesn't
  collapse the redundancy and silently break stale-parquet recovery.

BTN3A1 / butyrophilin gap tracked separately in #230.

* Speed up is_non_peptide_ligand opt-in test, tighten triple-derive comment

- tests/test_export.py: opt-in integration test now uses load_observations
  with a column projection ... (continued)

3948 of 5520 relevant lines covered (71.52%)

0.72 hits per line

Coverage Regressions

Lines Coverage ∆ File
130
79.9
 -0.07% export.py
36
89.3
 0.15% curation.py
14
79.43
 0.0% scanner.py
9
94.34
 -0.18% observations.py
2
88.75
 0.0% supplement.py
Jobs
ID Job ID Ran Files Coverage
1 25389171834.1 28
71.52
 GitHub Action Run
Source Files on build 25389171834