Researchers at the Icahn School of Medicine at Mount Sinai have created a gene set foundation model (GSFM) that learns how genes work together inside human cells by analyzing patterns from millions of published gene sets.
The model draws inspiration from large language models like ChatGPT, learning the “meaning” of genes through their biological context rather than through raw expression data alone. A single gene can play different roles depending on cellular conditions, much like a word carries different meanings depending on the sentence it appears in.
GSFM was trained on gene sets from hundreds of thousands of independent studies and can identify the function of poorly understood genes, highlight genes involved in disease processes, and suggest potential new drug targets and biomarkers. The model demonstrated strong performance in predicting gene-gene and gene-function relationships that were later confirmed experimentally.
The research, published in Patterns, is openly accessible at gsfm.maayanlab.cloud. Future plans include combining GSFM with language models to generate natural-language explanations of gene functions and integrating it with drug-focused AI models to predict drug-cell interactions.
