Predicting Functional Effects of Genetic Variants

About

This collaborative project brings together researchers from HPI, MIT, and the Broad Institute to develop AI systems that predict how genetic mutations affect voltage-gated ion channels—proteins critical for nerve signaling, heart function, and muscle movement. By analyzing protein structures, gene expression data, and clinical records, the team’s models determine whether mutations hyperactivate (gain-of-function) or impair (loss-of-function) these channels, with applications in precision medicine and agricultural science.

Key milestones include creating a curated database of over 3,400 ion channel mutations with functional labels, training classical machine learning models for initial predictions, and developing a quantitative framework to compare how different AI architectures interpret biological data. The framework has proven versatile beyond proteins, showing promise for analyzing medical records in healthcare collaborations. The team is now integrating cutting-edge protein language models to boost accuracy while preparing a user-friendly portal to visualize mutation effects on 3D protein structures.