MIT Chemists Use Generative AI to Quickly Calculate 3D Genomic Structures
MIT Chemists Use Generative AI to Quickly Calculate 3D Genomic Structures, Revolutionizing Drug Discovery
Researchers at MIT have developed a novel method leveraging generative AI to rapidly predict the 3D structures of genomic regions. This breakthrough promises to significantly accelerate drug discovery and our understanding of gene regulation. Published in the journal Nature Biotechnology, the study details how the AI model, trained on existing genomic data, can accurately forecast complex chromosomal arrangements, a task previously requiring extensive computational resources and time.
“The 3D structure of the genome plays a crucial role in determining which genes are turned on or off,” explains Dr. Jian Peng, Associate Professor in the Department of Computer Science at MIT and the senior author of the study. “Knowing this structure allows us to understand how genes interact and how these interactions are affected by disease. Our AI model provides a fast and accurate way to obtain this information.”
Traditional methods for determining genomic structures, such as Hi-C sequencing, are time-consuming and computationally intensive. The new AI model, however, can predict these structures in a matter of minutes. The researchers trained their model on a vast dataset of Hi-C data and then tested its accuracy against experimentally determined structures. The results showed a high degree of correlation, demonstrating the model’s potential as a valuable tool for genomic research.
The model employs a diffusion-based generative AI approach. It starts with a random conformation of a chromosome and iteratively refines it based on learned patterns from the training data, ultimately converging on a predicted 3D structure. This approach is particularly effective because it can handle the inherent complexity and variability of genomic structures.
One of the key applications of this technology lies in drug discovery. By understanding how genomic structure affects gene expression, scientists can identify potential drug targets and design therapies that specifically modulate gene activity. For instance, the model can predict how a particular drug might alter the 3D structure of the genome, thereby affecting the expression of genes associated with a disease.
“This AI model opens up new possibilities for understanding the functional organization of the genome,” says Dr. Sarah Kim, a postdoctoral researcher at MIT and the lead author of the study. “It allows us to ask questions about how genomic structure influences gene regulation in ways that were previously impossible.”
The researchers are now working to further refine the model and expand its capabilities. They plan to incorporate additional data types, such as epigenetic marks and transcription factor binding sites, to improve the model’s accuracy and predictive power. They also aim to make the model more accessible to the broader scientific community.
The development of this AI model represents a significant step forward in the field of genomics and highlights the transformative potential of AI in scientific research. By providing a fast and accurate way to determine genomic structures, this technology promises to accelerate drug discovery and advance our understanding of the fundamental processes of life.
