pirl-unc / tsarina / 24396444783
74%

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: 31.522% (+2.2%) from 29.365%
24396444783

push

github

web-flow 
v0.5.2: hitlist 1.6.0 — gene_name pushdown + mono-allelic filter (#3)

* v0.5.2: hitlist 1.6.0 — gene_name pushdown + mono-allelic filter

hitlist 1.6.0 adds a peptide_mappings sidecar and semicolon-joined
gene_names/gene_ids/protein_ids columns on observations.parquet, closing
hitlist#43. This PR threads the new API through tsarina.

Default `tsarina hits --gene PRAME` path drops ProteomeIndex entirely — one
`load_observations(gene_name=...)` call replaces the previous Ensembl build
+ peptide enumeration + scan chain. Multi-gene peptides (MAGE family, CTAGs)
now surface all members in output instead of silently collapsing to one
"best" gene.

Changes:

- cli_hits: default path uses `load_observations(gene_name=..., mhc_class=...,
  species=...)`. ProteomeIndex kept only for `--skip-ms-evidence` and the
  `--iedb`/`--cedar` raw-CSV override.
- cli_hits: new `--mono-allelic-only` flag filters to `is_monoallelic=True`
  rows (direct observation, no predictor-based deconvolution).
- cli_hits: `--serotype A24` now actually matches A*24:02 — hitlist#44
  currently mis-reports canonical serotype as Bw4 (A*24:02 carries the Bw4
  public epitope), so we fall back to locus-prefix matching until upstream
  fix lands.
- cli_hits: defensive empty-frame handling — no more KeyError on empty
  post-filter aggregates.
- indexing.load_ms_evidence: pushes `peptide=` filter to the parquet reader
  when supported (hitlist 1.6.0+), falls back to in-memory isin otherwise.

Dogfood: `tsarina hits --gene PRAME --allele "HLA-A*24:02"` returns 2 hits
in 1.5s (Python startup + cold query). `--mono-allelic-only` surfaces 11
high-confidence PRAME peptides.

Tests: +8 (mostly serotype/allele/resolution filter behavior). 142 passed.

* Fix --serotype filter to use mhcgnomes membership instead of substring

The previous prefix-match approach constructed `-A*2` from serotype name
A2 and looked for it in `HLA-A*02:01`, but the molecular form has a
leading-zero (`-A*02`) so the substri... (continued)

609 of 1932 relevant lines covered (31.52%)

0.32 hits per line

Coverage Regressions

Lines Coverage ∆ File
100
25.63
 25.63% cli_hits.py
Jobs
ID Job ID Ran Files Coverage
1 24396444783.1 30
31.52
 GitHub Action Run
Source Files on build 24396444783