pirl-unc / mhcgnomes / 33068458204
89%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 40
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 89.094% (+0.08%) from 89.013%
33068458204

push

github

web-flow 
Stop returning confident results for inputs that justify nothing (#108)

* Stop inferring an unjustified species for ambiguous inputs

Two independent defects made an ambiguous input resolve to a confidently
wrong species.

1. A species prefix is visible to every descendant, so "BoLA" matches
   Bos sp. and also Bubalus bubalis. All of them became candidates and
   the tie fell through to sort_key's final `str(result)` tiebreak, which
   is alphabetical on the repr. 'Bubalus' sorts after 'Bos', so
   parse("BoLA class I") returned water buffalo while
   parse("BoLA-N*01301") returned Bos sp. Eight prefixes were affected
   this way, including RT1 class I -> Rattus villosissimus and
   NHP class I -> Saimiri sciureus.

   Prefer the least specific species when results are otherwise
   indistinguishable: if a single string matched an ancestor and its
   descendants, nothing in the input justified a descendant. Explicit
   descendant prefixes (Bubu-DQA, Bota-DRB3*011:01) are unaffected.

2. An unprefixed gene symbol shared by several species picked the
   "best-characterised" one by Species.num_genes, which counts inherited
   genes. "Galliformes sp." defines BLB1/BLB2 for every galliform, so
   Coturnix japonica outranked Gallus gallus (52 vs 46 visible genes)
   even though quail never declares those genes and its own entry uses
   the Coja-DAB1/DBB1/DCB1 nomenclature. Bare BLB2*02 resolved to quail
   while bare BF2*02:01 resolved to chicken.

   Rank by the new Species.num_own_genes instead, which counts only the
   genes a species declares in its own ontology entry (chicken 24, quail
   22), so both bird genes now agree on chicken.

Fixes #103
Fixes #105

Verified against the bundled netMHCpan/netMHCIIpan/IEDB corpora: exactly
1 of 11,558 real allele names changes, "B12 class I", from
Macaca fascicularis to genus-level Macaca sp. -- the same correction.
No measurable parse throughput change.

Claude-Session: https://claude.ai/code/session_012s4siLj2... (continued)

4011 of 4502 relevant lines covered (89.09%)

0.89 hits per line

Coverage Regressions

Lines Coverage ∆ File
72
88.69
 0.18% species.py
59
94.24
 0.06% parser.py
Jobs
ID Job ID Ran Files Coverage
1 33068458204.1 40
89.09
 GitHub Action Run
Source Files on build 33068458204