v0.1.1

variant-confidence v0.1.1

Maintenance release: adds ESM-1v/EVE score integration (T14) plus always-on
GitHub Actions CI, without changing the existing AlphaMissense calibration
behaviour. Backwards compatible.

Verified in clean clone (no network):

• ruff check . → All checks passed!
• pytest tests/ → 38 passed
• GitHub Actions: green on Python 3.11 / 3.12, 0 Code Scanning alerts

Added

• ESM-1v / EVE integration (T14): variant_confidence/data/esm_eve.py loads
  user-converted (chrom,pos,ref,alt,score) TSVs for ESM-1v (MIT, Meta) and EVE
  (MIT, Pascal Notin). join_scores returns NaN (never 0) for unmatched, same
  pattern as AlphaMissense. integrate.align_scores_esm_eve wires it to the
  pipeline; weights are NEVER committed (Opción A). Licenses verified 2026-07-18.
• CLI wiring (T14g): variant-confidence --source {synthetic,alphamissense,esm1v,eve} with --score-path selects the score
  source end-to-end. Reuses align_scores / align_scores_esm_eve; fail/degrade
  stays centralized in run_calibration (AC12); the report declares source=
  and a missing score is never imputed as 0. +8 tests.
• CI (github-actions-sonarcloud): .github/workflows/ci.yml (ruff +
  pytest+cov on py3.11/3.12, bandit→SARIF to Code Scanning, SonarCloud gated on
  SONAR_TOKEN). sonar-project.properties, scripts/bandit2sarif.py.

Fixed

• T13b commit message said "9 new tests" — the diff added 2 (suite 26 → 28);
  corrected in CHANGELOG (v0.1.0 entry).
• AC13b coverage figure (19.118 proteins) marked as NOT independently audited.

Known issues

• AlphaMissense license ambiguity (CC BY 4.0 vs CC BY-NC-SA 4.0) remains
  unresolved — treat the data as restricted until clarified.
• The live AlphaMissense / ESM-1v / EVE join is exercised by an offline fixture,
  not a live download in CI (weights/predictions are never committed).

v0.1.0 — Calibrated confidence layer for variant-effect pathogenicity

variant-confidence v0.1.0

Calibrated confidence layer for protein variant-effect pathogenicity predictions (AlphaMissense / ESM-1v / EVE). Does not train a new model — adds an auditable calibration layer on top of existing predictors.

Verified in clean clone (no network)

• ruff check . → All checks passed!
• pytest tests/ → 28 passed in 8.90s

What's in 0.1.0

• Calibration (AC1, AC1b): Platt scaling, isotonic regression, and split/Mondrian conformal prediction.
• ECE metric (AC2, AC9): equal-width + adaptive bins, bootstrap CI, low-count bins flagged as low-reliability.
• Leakage-free split (AC3): temporal by ClinVar release date + gene isolation; index-alignment bug fixed (silent-liar class).
• Missing data (AC4, T13b): explicit missing-score handling + structured-missing detection.
• Non-deceptive reporting (AC7): CLI emits score + interval + method + ECE before/after.
• AlphaMissense join (T13): license-safe integration.
• Degenerate-ECE bug fixed: synthetic generator is now discriminative; AUC preserved after calibration (no collapse to base rate).

License

AGPL-3.0-or-later — Pedro Sordo Martínez (amurlaniakea@gmail.com), 2026.

Install

pip install variant-confidence
variant-confidence --help