openvax / varcode / 24674584933
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
24674584933

push

github

web-flow 
Add cryptic-exon candidate enumerator (PR 11; #252) (#334)

Pure-Python motif-based enumerator for candidate cryptic splice
sites and internal exons on SV-rearranged alleles. No external
ML models; designed as a *candidate enumerator* for downstream
scoring by SpliceAI / Pangolin / RNA-seq junction support /
long-read assembly.

``varcode/cryptic_exons.py``
  Consensus PWM scoring (Mount 1982, Shapiro & Senapathy 1987):

  * 5' donor: 9 bp window, ``MAG|GTRAGT`` consensus, strict GT
    anchor at +1/+2 (non-canonical = score 0).
  * 3' acceptor: 4 bp core, ``YAG|G`` consensus, strict AG anchor
    at -2/-1.
  * Scoring = position-match ratio. Deliberately crude; the point
    is to enumerate candidates, not be the final caller.

  ``enumerate_candidates(contig, sequence, sequence_start, ...)``
  scans a sequence for acceptor-donor pairs forming plausible
  internal cryptic exons, returning
  :class:`CrypticExonCandidate` instances sorted by summed motif
  score.

  ``enumerate_from_structural_variant(variant, ...)`` convenience
  wrapper: scans the reference flanking the SV breakpoints, OR
  scans ``variant.alt_assembly`` when the SV carries a long-read-
  resolved allele — grounding candidates in actual molecules
  rather than inferred reference+breakpoint reconstructions.

External-integration hooks (documented, not wired here):

* **External predictors** — ``score_fn=`` parameter accepts a
  callable with signature ``(window, kind) -> probability``.
  SpliceAI / Pangolin wrappers plug in here without varcode
  depending on their models.
* **RNA evidence** — :class:`CrypticExonCandidate` wraps as an
  :class:`Outcome` with ``source="isovar"`` and
  ``evidence={"junction_reads": N}`` alongside the varcode-
  motif nomination. Downstream consumers filter by source.
* **Long-read assembly** — SV callers that carry
  ``alt_assembly`` (from hifiasm / Shasta / local targeted
  assembly) get candidates nominated on the assembled allele.

Tests: 12 cases co... (continued)
Source Files on build 24674584933
Detailed source file information is not available for this build.