AI Model Maps Unique Biological Pathways in Individuals

We all know individuals react differently to the same stimulus, from medication to exercise, but pinpointing the precise biological reasons has been a long-standing challenge. Now, researchers have introduced a new computational tool, the 'bioreaction-variation network', to illuminate these hidden differences.

This tool is a specific type of AI known as a graph neural network (GNN), expertly designed to understand complex relationships within a network. To teach it the language of muscle biology, the team trained it on a massive dataset of approximately 65,000 published scientific studies.

When tested with real gene expression data from mice subjected to exercise, the model successfully inferred individualised biological maps. These networks revealed not only the pathways common to all mice but also the unique molecular routes specific to each one. This work serves as a powerful proof of concept for creating customised, context-based models to explain biological variation at a personal level.