Large Genome Model: Open-Source AI Trained on Trillions of DNA Bases

Summary

Ars Technica reports on a new open-source AI model trained on trillions of DNA base pairs from massive genomic datasets. Building on earlier work such as the Evo system, which was trained on bacterial genomes, this large genome model can identify genes, regulatory sequences, splice sites, and other biologically significant features directly from raw sequence data. The model represents a significant scaling effort in applying transformer-based architectures to biological sequences rather than nat