pirl-unc / hitlist / 25011520130

Build # 25011520130

Build Type

push

github

Committed by web-flow

Commit Message

v1.23.0: Exact-allele bag expansion + provenance for class-only rows (#137) (#163)

Adds three columns to every observation row so multi-allelic / class-only
MS rows become usable training data instead of being dropped:

- mhc_allele_set: ";"-joined candidate 4-digit alleles
- mhc_allele_provenance: exact | sample_allele_match | pmid_class_pool | unmatched
- mhc_allele_bag_size: int

Resolution policy
- four-digit row -> bag = {row's allele}, provenance "exact"
- class-only + IEDB "Host | MHC Types Present" -> filter pool to row's
  mhc_class, provenance "sample_allele_match"
- class-only with no per-row donor genotype -> fall back to PMID's curated
  hla_alleles pool (flat list / dict-of-lists / dict-of-strings all
  flattened to 4-digit strings), provenance "pmid_class_pool"
- anything else (two-digit, serological, unresolved without pool match) ->
  empty bag, provenance "unmatched"

Downstream filters
- generate_observations_table / generate_binding_table accept
  mhc_allele_provenance=... and mhc_allele_in_bag=... so callers can pick
  exact-only (strict point-label training) or exact + sample_allele_match
  (MIL / noisy-OR over small trusted bags).
- CLI: --mhc-allele-provenance and --mhc-allele-in-bag on `obs` and `bind`.

Wires the same expansion through hitlist.scanner (IEDB ingest reads index
49 "Host | MHC Types Present") and hitlist.supplement (per-allele
classification reuses the per-PMID curated pool when supplements omit
donor genotype).

Coverage Stats

2628 of 4292 relevant lines covered (61.23%)

0.61 hits per line

Source File

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

Source Not Available