pirl-unc / tcrsift / 25881973801

Build # 25881973801

Build Type

push

github

Committed by web-flow

Commit Message

Funnel + QC + signature batch: closes #72/#74/#75/#76/#77 (#78)

Five tightly-related plotting/QC issues bundled. Each is small on its
own; they share the new `tcrsift.format` and `tcrsift.candidate`
modules and the extended funnel/QC primitives.

#77 — Pretty method labels
- New module `tcrsift/format.py` with `pretty_method`, `pretty_methods`,
  `pretty_sample`, `pretty_samples`. Translates `AIMpos` → `AIM⁺`,
  `CTYneg` → `CTY⁻`, `CTYneg_tetpos` → `tet⁺CTY⁻` (CTY-second
  reordering for two-part names).
- Applied in `plot_method_overlap` (heatmap axes), `plot_method_recovery`
  (xticks), `plot_clone_freq_vs_signature_per_sample` (panel titles),
  the rank curves, and the top-N labeled bars.

#74 — Per-clone signature scores in clonotypes.csv
- New `compute_signature_scores_per_clonotype(per_cell_df, ...)` in
  `gex.py`. Returns one row per clonotype with
  `signature_antigen_response`, `signature_cytolytic`,
  `signature_tumor_reactive` columns (plus any other signatures the
  caller passes via `signatures=`).
- Mean-log1p across each gene set, then group-mean per clonotype.
  Restricts to CD8+ cells by default (auto-detected via
  `gex.CD8A`/`gex.CD8`); `cd8_only=False` opts out.
- Wired into `cmd_annotate_gex` so `tcrsift annotate-gex --aggregate`
  emits the columns automatically when per-cell augmentation ran.

#75 — Signature-based candidate selection
- New module `tcrsift/candidate.py` with `select_candidates(clonotypes,
  signatures, top_n)`. Adds boolean tracking columns:
  `tier1_or_tier2`, `tier3plus_top_by_{name}_signature` per signature,
  `is_selected`. Pure: doesn't filter or mutate input.
- Selection: tier1 ∪ tier2 ∪ top-N-per-signature(tier3 ∪ tier4 ∪ tier5).
  Tier1/2 don't consume top-N slots in the tier3+ pool. NaN scores
  sort to the bottom and don't get picked.

#76 — Stacked cells-per-sample QC bars
- New `plot_cells_per_sample_stacked(adata, output_path)` renders
  three layered bars per sample: loaded barcodes (pale gray) /
... (continued)

Coverage Stats

5913 of 7801 relevant lines covered (75.8%)

3.03 hits per line

Source File

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

Source Not Available