Evo 2: A Large-Scale Genomic Model

In 2025, an AI system called Evo was introduced, trained on a massive number of bacterial genomes. This system was able to identify the next gene in a cluster of related genes or suggest a completely new protein.

The first version of Evo worked because bacteria tend to cluster related genes together. However, this is not true for organisms with complex cells, which have equally complex genome structures.

Evo 2 represents a significant evolution. This model has been trained on genomes from all three domains of life (bacteria, archaea, and eukaryotes). After training on trillions of base pairs of DNA, Evo 2 developed internal representations of key features even in complex genomes like ours, including regulatory DNA and splice sites, which can be challenging for humans to spot.