AI Model Predicts Gene Activity, Revolutionizing Biology: Columbia University's GET

Columbia University researchers developed an AI model, General Expression Transformer (GET), capable of predicting gene expression in cells. This advancement aims to broaden the understanding of genetic diseases and cancer, paving the way for cell-specific gene therapies. GET was trained using an approach similar to that of language models like ChatGPT. While ChatGPT learns language grammar, GET learned the underlying rules for genes: how they are activated or deactivated and how their activity is regulated. This ability to predict gene expression is crucial, as proteins play a role in nearly all biological processes. Raul Rabadan, a study author and director of the Program for Mathematical Genomics at Columbia University, stated that "biology is transforming into a predictive science," potentially marking "a revolution in biology." Data used to train GET came from over 1.3 million cells, covering 213 different cell types in the human body. Mark Gerstein, Professor of Biomedical Informatics at the Yale School of Medicine, noted that experts have been attempting to make predictions about gene regulation for years. A notable aspect of GET is its ability to make predictions about cell types not directly included in the training. Researchers hope the model will aid in developing gene therapies that correct mutations damaging specific cell types. The ability to predict which genes are turned on or off in different cells could help determine a disease's cell of origin. GET could also facilitate decision-making in research. In cancer research, where numerous mutations often exist in the genome, GET could help identify relevant genetic combinations, streamlining research efforts.