AI Helps Decode Gene Interactions in Disease

Researchers from Gladstone Institutes, the Broad Institute of MIT and Harvard, and Dana-Farber Cancer Institute utilized artificial intelligence (AI) to gain insights into the complex interplay of genes and their role in disease. In a study published in Nature, Dr. Christina Theodoris and her team introduced Geneformer, an AI model that uses transfer learning to understand gene interactions and predict disruptions that lead to disease.

Gene networks, consisting of interconnected genes that regulate cellular function, have long been challenging to map due to their complexity. However, Geneformer helps to overcome this hurdle with a foundation of knowledge acquired from vast amounts of data on gene interactions across a wide range of human tissues.

For their work, the team showcased Geneformer's capabilities in studying heart disease. By analyzing cardiomyocytes, the team successfully identified genes that play a role in heart disease. Geneformer also uncovered previously unknown genes associated with heart disease, such as TEAD4. Removing TEAD4 from cardiomyocytes in the lab decreased the cells' beating ability, supporting Geneformer's predictions and its potential to uncover insights into disease mechanisms.

The transfer learning approach employed by Geneformer proved particularly valuable in situations with limited data. By pre-training the model with a comprehensive dataset of gene activity in millions of cells, Geneformer gained a fundamental understanding of gene interactions. This allowed it to make accurate predictions even with minimal examples of relevant data.

Dr. Theodoris and her colleagues also highlighted the potential of Geneformer in drug target discovery. By focusing on genes that play a central role in gene networks, rather than peripheral ones, Geneformer could influence new therapeutic strategies.

The researchers plan to expand Geneformer's analysis to include more cell types, further enhancing its ability to decipher gene networks. Additionally, they have made the model open-source, empowering other scientists to utilize it for their own investigations. These findings not only advance our understanding of gene networks but also offer a promising avenue for accelerating the development of targeted therapies.