ashkankhilwatgar1/Variant-interpretation-UI

HHT ACMG Variant Classifier — Demo

Browse ACMG/AMP + ClinGen HHT-VCEP (GN135/GN136) variant classifications for ~201 pre-computed variants. Every criterion is a checkbox you can toggle, variable-strength criteria get a strength selector, and the classification recalculates live in your browser using scoring.js — a line-for-line JS port of the pipeline's tools/scoring.py classify(), verified against it on all 201 cached variants plus synthetic edge cases (see scoring.js's header comment).

This is a fully static site (HF Spaces sdk: static): no backend, no Python, no API keys. Cached results are plain JSON files fetched client-side.

Live classification (running a new, arbitrary variant through the full LLM + genomics-API pipeline) is disabled in this demo — there is no server here to run it even if it weren't. The "Run pipeline" button is visibly disabled and shows a notice directing you to contact akhilwat@hamilton.edu for live-run access.

Files

• —index.html, styles.css, app.js — the GUI (adapted from the original Flask-backed demo to fetch static JSON instead of calling API routes).
• —scoring.js — ported classification logic (see above).
• —meta.json — criterion metadata (automatable / manual / excluded lists, descriptions, thresholds) — a one-time static dump of the original data/planrag.py's PLANRAG_DB / EXCLUDED_CRITERIA.
• —scoring_rules.json — the combining-criteria rules (fixed/variable strength buckets, incompatible combinations, classification rule tables) — a one-time static dump of PLANRAG_DB's SCORING block, consumed by scoring.js.
• —data/results/*.json — one file per cached variant (pre-populated criterion applies/strength/evidence state), plus data/results/index.json listing all available variants for the dropdown.