pirl-unc / hitlist / 24905543890

Build # 24905543890

Build Type

push

github

Committed by web-flow

Commit Message

v1.16.0: Line-expression anchors + peptide-origin join (#140) (#145)

* v1.16.0: Line-expression anchor layer with 6-tier fallback (#140)

Add a curated per-line RNA/transcript anchor registry so sample-aware
training exports can attach expression evidence to peptide observations
while preserving full provenance (issue #140):

- hitlist/data/line_expression_anchors.yaml: 21 line entries covering
  the issue-#140 starter list (C1R, 721.221, T2, JY/HHC/GR/CD165/RA957,
  HAP1, HeLa, HEK293, THP-1, SaOS-2, A375, K562, GM12878) plus common
  engineered derivatives (HeLa.ABC-KO, C1R-HLA, HAP1-KO, etc.).
- hitlist/data/line_expression/: per-source metadata + a packaged
  GM12878 gene-TPM subset from ENCODE (~200KB, 23.7K genes).
- hitlist/line_expression.py: 6-tier resolver (exact / parent /
  family / cancer-type / tissue / none) with placeholder-aware
  tier-1 guard so entries without shipped data correctly fall
  through. Also exposes the peptide-origin compute kernel, which
  supports transcript-isoform-aware TPM summing via an injectable
  lookup (so tests exercise the summation logic without pyensembl).
- hitlist/builder.py: build_line_expression emits line_expression.parquet
  from packaged CSVs + optionally-registered DepMap matrices with an
  atomic-rename write and the harmonized source-metadata schema.
- hitlist/downloads.py: register DepMap gene / transcript TPM as
  manual downloads.
- hitlist/export.py: generate_sample_expression_table per-sample
  anchor dump; with_peptide_origin flag on generate_training_table
  that attaches expression_backend/expression_key/expression_match_tier/
  expression_parent_key and peptide_origin_gene/tpm columns via a
  one-TPM-load-per-line + cached per-(peptide, line) scorer.
- hitlist/cli.py: hitlist export samples --with-expression-anchors,
  hitlist export training --with-peptide-origin, and a new
  hitlist export line-expression subcommand.
- tests/test_line_expression.py (25 tests) covering every tier and
 ... (continued)

Coverage Stats

2289 of 4032 relevant lines covered (56.77%)

0.57 hits per line

Source File

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

Source Not Available