Open Source AI Trained on Trillions of Genetic Bases Can Identify Genes and Regulatory Sequences Across All Life

Researchers have released Evo 2, an open-source large genome model trained on trillions of DNA bases that can identify genes, regulatory sequences, splice sites, and other functional elements across all domains of life. The model represents a new frontier in computational biology, applying the same transformer architecture that powers language models like ChatGPT to the four-letter alphabet of DNA. By learning the patterns of genomes from bacteria to humans, Evo 2 can predict genetic function and design novel sequences with a breadth that would take human researchers decades to achieve manually.