pirl-unc / hitlist / 25068532310
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 25
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: 68.658% (+0.03%) from 68.625%
25068532310

push

github

web-flow 
v1.29.8: pmhc table polish — normalize alleles, sort by evidence, header-once (#180)

* v1.29.7: Replace "exploding" jargon in pmhc progress messages

User asked what "exploding gene-list columns" meant — pandas jargon
leaking into a CLI status line. Rephrased to plain English:

  before: "exploding gene-list columns..."
   after: "splitting multi-gene rows (one row per gene)..."

The underlying operation is pandas' ``DataFrame.explode`` — necessary
because each observations row stores ``gene_names`` as a semicolon-
joined string (one peptide can map to multiple genes), and we need
one row per (gene, allele, peptide) for the grouping. Comment near
the call still mentions ``DataFrame.explode`` so readers tracing the
code can find the API.

* v1.29.8: pmhc table polish — normalize alleles, sort by evidence, header-once, --predictor tip

User feedback against v1.29.6 PRAME output called out four polish issues:

1. **Sort allele groups by total evidence count desc**, not alphabetically.
   The most-attested allele for a gene is the one users care about
   first; alphabetical "A*02:01 → A*68:02 → ..." buried it.

2. **Normalize MHC restriction before grouping**. The parquet stores
   both ``A*02:01`` and ``HLA-A*02:01`` for the same allele because
   different sources used different conventions; passing the raw
   strings to groupby split peptides across two unrelated buckets
   (PRAME's SLLQHLIGL was reported as 1 obs on ``A*02:01`` AND 15 obs
   on ``HLA-A*02:01``). ``curation.normalize_allele`` (mhcgnomes-backed)
   collapses them.

3. **Print column headers once per gene, not per allele**. The
   peptide / n_obs / pmids header was repeated under every allele
   group; for PRAME's ~30 alleles that's a wall of redundant text.
   Now printed once at the top of each gene section, with peptide
   rows beneath each allele line lining up under the same column rule.

4. **Append a tip about ``--predictor netmhcpan``** when no predictor
   was passed, since use... (continued)

3229 of 4703 relevant lines covered (68.66%)

0.69 hits per line

Coverage Regressions

Lines Coverage ∆ File
19
77.33
 0.78% pmhc_query.py
Jobs
ID Job ID Ran Files Coverage
1 25068532310.1 25
68.66
 GitHub Action Run
Source Files on build 25068532310