pirl-unc / tsarina / 25756991701

Build # 25756991701

Build Type

push

github

Committed by web-flow

Commit Message

Bump hitlist pin to >=1.30.49; re-attach gene columns post-1.30.46 (#73)

* Bump hitlist pin to >=1.30.49 and fix gene-column attach

Three changes, all driven by an audit of hitlist 1.15.1 → 1.30.50 for
runtime impact on tsarina.

1. Bump `hitlist>=` pin in pyproject.toml from 1.15.1 to 1.30.49. The
   prior floor was decoupled from what tsarina actually consumes:

   - `mhc_allele_set` / `mhc_allele_provenance` columns (read by
     spanning.py) arrived in v1.23.0.
   - `peptide_attribution` provenance handling (commit 28fe1c5)
     requires v1.30.39.
   - `ProteomeIndex.from_ensembl` on-disk cache (the one we've been
     waiting for) lands in v1.30.49. Closes the third bullet of #69 —
     per-patient `tsarina personalize` runs in shell loops now skip
     the cold ~10-60 s Ensembl walk after the first invocation.

2. Re-attach gene_names / gene_ids / protein_ids in
   `cli_hits._handle_cached_path` after a no-projection
   `load_observations` / `load_all_evidence`. hitlist v1.30.46 split
   those columns out of observations.parquet (#238) — they're now only
   auto-derived when the caller explicitly lists them in `columns=`.
   tsarina's fast path didn't project, so on post-1.30.46 parquets the
   downstream `gene_ident_cols` merge silently dropped gene
   identifiers from `peptides` / `pmhc` / `refs` outputs. Re-attach via
   hitlist's own `annotate_observations_with_genes` against the
   peptide_mappings sidecar so behavior matches pre-1.30.46 parquets
   (which still carry the columns natively) and post-1.30.46 (which
   don't). Update the refs-handler test that was asserting the old
   "no gene_ident_cols passthrough" invariant; add a fresh regression
   test pinning the attach behavior.

3. Update docstrings on `human_exclusive_viral_peptides` /
   `cancer_specific_viral_peptides` to reflect that hitlist 1.30.49+
   persists the proteome index to disk (~/.hitlist/proteome_index_cache),
   so the cache is no longer just process-local.

Refs #... (continued)

Coverage Stats

2397 of 3387 relevant lines covered (70.77%)

0.71 hits per line

Source File

Press 'n' to go to next uncovered line, 'b' for previous

Source Not Available