openvax / vaxrank / 26072242668
96%
master: 88%

LAST BUILD BRANCH: v3.1.8
DEFAULT BRANCH: master
Ran
Jobs 0
Files 0
Run time –
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
26072242668

push

github

web-flow 
WT prediction: use predict_from_named_peptides, name by mutant peptide (#301)

* Use predict_from_named_peptides for WT, name entries by mutant peptide

The WT prediction call passes already-extracted peptides, not protein
fragments to scan — so the k-mer-sliding behavior of
predict_from_named_sequences was a hazard, not a feature. It forced a
dedup on WT-peptide strings plus a (peptide, allele) lookup key, and
required synthetic "wt_N" names that carried no information.

Switching to predict_from_named_peptides scores each peptide exactly
once, which lets us name each WT entry after the mutant peptide it
pairs with. The lookup at the consumer site then collapses from a
two-hop (mutant -> wt_peptides[mutant], then (wt_peptide, allele))
to a one-hop ((mutant_peptide, allele)).

* Address review: trim comment, add focused WT-path test

- Shrink the WT-predict comment to 3 lines (down from 6); the
  no-k-mer-sliding half-sentence carries the load.
- Add test_wt_prediction_uses_named_peptides_keyed_by_mutant: stubs
  both predict_from_named_sequences and predict_from_named_peptides,
  asserts (a) the WT call's input dict is keyed by the mutant peptide
  string, and (b) the resulting CandidateEpitope's wt.sequence comes
  from the prediction row's own ``peptide`` column — locking in the
  one-hop lookup invariant.

(The third review item — tightening the topiary pin — turns out to
be moot: 5.16.0 doesn't exist on PyPI, so ``>=5.16.0`` already
resolves to 5.16.1+, both of which have predict_from_named_peptides.)

* Compat: drop topiary 5.16.2's empty-string sample_name column

Topiary 5.16.2 now emits a ``sample_name`` column populated with empty
strings on every predict_* call. Its EvalContext groups by
``sample_name`` as the leading level whenever the column is present
with any non-NaN value — and empty strings are not NaN, so the
group_index silently inflates from 4 levels to 5. Vaxrank's lookup
keys are 4-tuples and the DSL consumer unpacks 4 elements p... (continued)
Source Files on build 26072242668
Detailed source file information is not available for this build.