snv-judge

What problem does it solve?

SNV-judge provides calibrated, ACMG-aware pathogenicity predictions for human missense single-nucleotide variants (SNVs) by integrating 8 heterogeneous features from free public sources, enabling clinicians and researchers to interpret variants with quantitative probabilities and actionable classifications.

Core Features & Use Cases

• Calibrated probability of pathogenicity (0–1) and ACMG 5-tier classification (P/LP/VUS/LB/B) for missense SNVs.
• Per-feature SHAP contributions with visual explanations, supporting transparent interpretation.
• Optional clinical interpretation reports generated via LLM templates (Chinese/English/summary).

Quick Start

Provide a chrom/pos/ref/alt or protein change and run predict_variant to obtain a calibrated pathogenicity probability, ACMG classification, and SHAP explanations.