AI-Accelerated 3D Genome Structure Prediction

MIT chemists have developed a new generative AI model, ChromoGen, to rapidly predict the three-dimensional structures of genomes within individual cells.

This method significantly accelerates the process, predicting thousands of structures in minutes compared to the days required by existing experimental techniques like Hi-C.

ChromoGen analyzes DNA sequences and chromatin accessibility data to generate multiple possible 3D structures, reflecting the inherent variability of DNA conformation.

The model's accuracy has been validated by comparing its predictions to experimentally determined structures, showing strong agreement and applicability to various cell types.

This advancement opens new avenues for researching how 3D genome organization influences gene expression, cell function, and disease, with the model and data made publicly available.

 

 https://news.mit.edu/2025/with-generative-ai-mit-chemists-quickly-calculate-3d-genomic-structures-0131