Evo 2 AI: Predicting and Engineering Genetic Sequences at Unprecedented Speed

Generative AI Tool Transforms Biology and Boosts Life Sciences Research

A recently created generative AI tool, Evo 2, has the potential to greatly influence the future of biology. Developed in collaboration with a team led by Stanford assistant professor Brian Hie, Evo 2 allows researchers to forecast and engineer genetic sequences with speed, propelling the discovery of new biological functions and innovations. The technology has the ability to reframe the study of human disease and health, bioengineering, and disease.

Evo 2 makes use of a dataset so massive and wide-ranging in life form—even including humans, plants, bacteria, and extinct species—that it can simulate in just a few minutes what nature would only take years or even centuries to produce: entirely novel sequences of DNA. The tool can model interactions between genes, produce new sequences with specific desired functions, and, in doing so, could have some applications to medicine and environmental science.

The generative aspect of Evo 2 is similar to AI applications such as ChatGPT. Similar to how ChatGPT will predict the next word in a sentence, Evo 2 predicts the next nucleotide in a DNA sequence. By entering part of a gene sequence, Evo 2 foretells the rest of it and even predicts mutations that would provide new, useful characteristics. This predictive power is essential for researchers seeking to discover the genetic cause of disease and health, like finding mutations that result in cancer or other illnesses.

Evo 2 is also better than its previous version, Evo 1. While Evo 1 focused primarily on primitive organisms, Evo 2 has genomes of over 15,000 organisms, ranging from plants and animals to human beings. The model now accounts for almost 9 trillion nucleotides, giving a more detailed picture of biology. The increased dataset is essential in the creation of new genetic sequences that might be used to cure diseases or new bioengineering tools. Evo 2 also maintains safety by not including viral genomes, so as not to create dangerous pathogens.

The collaborative effort behind Evo 2 involved researchers from Stanford, NVIDIA, and the Arc Institute. These teams worked together to develop the machine learning algorithms, validate the model with biological expertise, and conduct real-world experiments to test the predictions. The AI tool has already been used to identify genetic variations that could lead to diseases, marking a promising step toward clinical applications.

In the future, Evo 2 can play a pivotal role in finding new genetic sequences with specific functions. The tool can accelerate genetic research and simplify the process of gene editing, such as CRISPR technologies. With continued development, the ability of the tool can be further incorporated with systems biology models to gain further insight into gene interactions and disease processes.

Source: Stanford University News, 2025.