openvax / varcode / 25587269410
88%

DEFAULT BRANCH: main
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
25587269410

push

github

web-flow 
Wire germline-aware effect prediction end-to-end (#268) (#360)

Treats germline as a transcript modifier, not a wrapper around
existing effects (per the design pivot proposed in #268's design-
update comment). One unified PR covering everything that was
previously sliced into 7 sub-issues — slice 2 (window-based
application), slice 3 (annotator dispatch), slice 4 (phase
enumeration → multi-outcome), slice 5 (LOH), slice 6 (splice
recomputation, mostly for free), with sub-1 (#282) collapsing
to just the PhaseHypothesis dataclass since GermlineAwareEffect
went away under the pivot. Slice 7 (docs + corpus) is the only
piece deferred — to a follow-up.

Mechanism
---------

``predict_germline_aware_effect(somatic, transcript, germline_ctx,
annotator, phase_resolver=None)`` is the single entry point. It:

1. Looks up germline variants in the somatic's window (default:
   the codon containing the somatic, expanded to a splice-signal
   window when splice-adjacent).
2. If no germline overlaps: delegates to the annotator unchanged.
   SPARSE / HOTSPOTS_ONLY contexts mark the result with
   ``germline_unknown=True`` so consumers see the uncertainty.
3. Detects LOH (somatic shares position+alt with a germline het call)
   and sets ``effect.is_loh = True``.
4. Resolves phase between somatic and germline-in-window via
   ``phase_resolver``. Hemizygous chromosomes (chrM, chrY) →
   automatic cis. PS-tagged + resolver answer → single hypothesis.
   Otherwise → enumerate up to 2^n hypotheses (capped at 8).
5. For each hypothesis: build patient baseline (just cis germline
   applied) + post-somatic haplotype (cis germline + somatic) →
   classify the diff using the protein-diff classifier. Same
   machinery the protein_diff annotator uses, just against the
   patient baseline instead of the reference.
6. Single hypothesis → return the classified effect directly.
   Multiple hypotheses → wrap in PhaseAmbiguousEffect (a
   MultiOutcomeEffect subclass; sibling of Haploty... (continued)
Source Files on build 25587269410
Detailed source file information is not available for this build.