26924421913

Committed 04 Jun 2026 01:34AM UTC coverage: 79.802% (+0.1%) from 79.665%

Build # 26924421913

Build Type

push

github

Committed by

web-flow

Commit Message

Per-condition signature-consistency QC: does each sort behave like its label (#161) (#164)

The single most diagnostic cohort signal — surfaced cheaply from the GEX
signatures tcrsift already computes.

- sort_qc.sort_signature_consistency(df, ...): per-(donor, sort) check.
  Within-donor: each sort's expected signature(s) are enriched (z-scored
  across the donor's cells) vs baseline; flags a sort that isn't. Cross-
  donor: a donor whose per-(sort, signature) matrix correlates poorly with
  the leave-one-out cohort consensus is flagged (the B1-4 signal — verified:
  consistent donors corr 0.99, the broken one -0.55 -> outlier).
- Config-driven sort->signature map; biology NOT hardcoded. NB CTYneg is a
  Cell Trace Yellow dilution gate (divided/expanded) -> Proliferation, not
  low cytolytic. Compound sorts (AIMpos_CTYneg) require all components;
  unmapped sorts (tetpos) reported but not flagged.
- sort_signature_consistency_from_adata: scores the needed signatures per
  cell via the shared expression resolver and runs the check.
- run pipeline emits sort_consistency_qc.csv + ⚠️ per flagged (donor, sort);
  best-effort (skips when GEX/signature genes absent). 12 new tests.

Bump 2.29.0.

Coverage Stats

8850 of 11090 relevant lines covered (79.8%)

3.19 hits per line

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

56.38

/cli.py

Source Not Available

The file "cli.py" isn't available on github. Either it's been removed, or the repo root directory needs to be updated.

pirl-unc / tcrsift / 26924421913

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

Source Not Available

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