MIT researchers have developed ChromoGen, a groundbreaking AI model that swiftly predicts the 3D structure of the genome, a task that previously took weeks, now achievable in minutes. This innovation, detailed in Science Advances, combines deep learning with generative AI to analyze DNA sequences and chromatin accessibility, enabling rapid generation of thousands of chromatin structures.
The 3D structure of the genome dictates gene expression and cellular diversity. ChromoGen uses a two-part system: one part analyzes DNA sequences, and the other, a generative AI component, predicts 3D structures based on a training set of over 11 million chromatin conformations. This allows researchers to study how the 3D genome organization affects gene expression and cell functions more efficiently.
Led by MIT Associate Professor of Chemistry Bin Zhang, the team's AI model can predict thousands of chromatin structures in minutes, significantly faster than current experimental techniques. The model's predictions align with experimental data and accurately predict data from other cell types, making it a valuable tool for exploring chromatin variations and their functional effects. The researchers have made their model and data publicly available, inviting further scientific exploration.