2
82%
main: 82%

Ran 03 Jun 2026 01:40AM UTC

Files 36

Run time 1s

Badge

Embed ▾

Committed 03 Jun 2026 01:37AM UTC coverage: 80.834% (+0.1%) from 80.695%

Job # 26858451065.2

Build Type

push

github

Committed by

web-flow

Commit Message

Composable in-silico filter layer (#149) (#155)

A stackable filter layer that composes on top of the assay groups and the
selection rules: each layer is a per-clone percentile predicate on Ppost
(#143) or a GEX signature score (#141/#142/#144), so a group is
progressively refined (CTYneg -> CTYneg.Ppost^low.GEXnaive^low...).

New tcrsift/insilico_filter.py:
- FilterPredicate {score, direction(low/high), percentile, rank_method}
  + predicates_from_config.
- percentile_rank: lower-is-better rank in [0,1], NaN-safe, ties grouped
  (Ppost==0 kept as a valid best value, never dropped).
- insilico_mask (pass all predicates; ranks computed within each assay
  group), average_percentile_rank (the composite's ranking metric).
- apply_insilico_filter: pass flag + avg rank + named composite '<grp>+IS'.
- expand_insilico_twins: each assay group gains a filtered twin (N -> 2N).
- insilico_overlap_long: Jaccard overlap across base groups AND their +IS
  twins, reusing compute_sample_overlap_matrices.

Config: free-form insilico_filter: block (label + ordered predicates),
parsed alongside selection. Default empty = off. Docs page + nav.

Bump 2.24.0.

Coverage Stats

8157 of 10091 relevant lines covered (80.83%)

0.81 hits per line

pirl-unc / tcrsift / 26858451065 / 2
82%
main: 82%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job python-3.11 - 26858451065.2

pirl-unc / tcrsift / 26858451065 / 2 82% main: 82%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job python-3.11 - 26858451065.2

pirl-unc / tcrsift / 26858451065 / 2
82%
main: 82%

README BADGES
x