ACMG Assistant is a student-level, research-oriented variant classification tool developed to explore the practical application of ACMG/AMP 2015 and 2023 guidelines. It combines automated retrieval of annotation data from public APIs with structured interactive evidence collection to support systematic variant interpretation.

Local-first biomedical MCP over AlphaFold DB + 8 public sources, with SQLite knowledge graph, offline mode, and explicit clinical-use limits.

Calibrate functional/in-silico variant scores into ACMG clinical evidence strengths (ClinGen-SVI), with LDLR/AlphaMissense worked example

Variant pathogenicity classifier with calibrated probabilities, gene-grouped cross-validation

Machine-learning triage of LDLR variants of uncertain significance using predictor concordance and ACMG-aligned evidence mapping

Research probe: LLM-assisted clinical variant classification with retrieval bundles. 60 ClinVar variants, no training.

ML model to predict pathogenicity and functional effects of genetic variants using ClinVar annotations and pre-computed scores

Machine-learning triage of LDLR variants of uncertain significance using predictor concordance and ACMG-aligned evidence mapping

Interpretable ML pipeline (Random Forest / XGBoost / SVM with SHAP) that predicts Poor vs. Intermediate CYP2C9 metabolizer phenotype for diplotypes of indeterminate function, and prioritizes high-impact variants for MD simulation and clinical review.

Privacy-preserving hybrid RAG pipeline for clinical genomic variant interpretation — local LLM, LangGraph, pgvector, ClinVar, gnomAD, ClinGen, NCCN

TF-DFE: A Topo-Fractal Dynamic Fuzzy Ensemble for pathogenicity prediction. Integrates genomic fractals (FCGR) & topological structures (TDA) to achieve 94.2% Accuracy and 0.886 MCC on 207K variants via selective abstention.