Researchers around the world now have access to Evo 2, a groundbreaking large artificial intelligence model capable of interpreting the genetic code of life. The largest publicly available genomic AI model to date was developed jointly by the Arc Institute and Stanford University on Nvidia’s DGX Cloud platform, with support from Amazon Web Services (AWS).
The start of a new era in genetic research
Evo 2, a large and pioneering artificial intelligence model capable of analysing and interpreting the genetic code of life, is now available to researchers globally. The largest publicly available genomic AI model to date was developed by the Arc Institute and Stanford University on Nvidia’s DGX Cloud platform, with support from Amazon Web Services (AWS).
Availability and how it works
Evo 2 is available worldwide through the Nvidia BioNeMo platform, and it can also be used as an Nvidia NIM microservice, which enables simple and secure deployment. The model was built by processing nearly 9 trillion (10¹²) nucleotides, the building blocks of DNA and RNA, and it can be applied in many areas of scientific research. For example, it can help predict proteins based on genetic sequences, identify new molecules for healthcare and industry, and assess the functional effects of gene mutations.
New dimensions of generative genomics
According to Patrick Hsu, co-founder of the Arc Institute and professor of bioengineering at the University of California, Berkeley, Evo 2 is a huge step forward in generative genomics. By gaining a deeper understanding of genetic building blocks, solutions to health and environmental challenges may emerge that were previously unimaginable.
The model can also generate various biological sequences and can be fine-tuned by researchers by adjusting its parameters. Those who want to develop it further on their own datasets can download the open-source version of the Nvidia BioNeMo Framework, which contains computing tools optimised specifically for biomolecular research.
A wide range of applications
Evo 2 can analyse DNA, RNA and protein data and can be used in many scientific fields, including healthcare, agricultural biotechnology and materials science. Thanks to its unique model architecture, it can process up to one million genetic sequence tokens, enabling scientists to explore the connections between the genetic code and cell function more deeply.
The model could be especially useful in drug research, as it can identify gene variants associated with a particular disease. According to research by the Arc Institute and Stanford University, for the BRCA1 gene, which is linked to breast cancer, Evo 2 can predict with 90% accuracy how a previously unknown mutation will affect the gene’s function.
In agricultural biotechnology, Evo 2 could help solve global food problems, for example by developing crop varieties that are more resistant to climate change or richer in nutrients. It can also be applied in other scientific fields, such as developing biofuels or creating proteins that can break down oil or plastic.
The limits of a new scientific tool
“Launching Evo 2 is like sending a powerful new telescope to the farthest reaches of the universe. We know it holds enormous potential, but we have no idea yet what discoveries await us,” said Dave Burke, chief technology officer of the Arc Institute.
AI accelerates innovation
The Arc Institute was founded in 2021 with an endowment of 650 million dollars to focus on long-term scientific challenges. One of the institute’s key goals is to free researchers from the constant pressure of applying for grants and give them room to innovate. Arc researchers work in eight-year, renewable funding cycles, which they can hold alongside their university positions. Its partners include Stanford University and the University of California (Berkeley and San Francisco).
Thanks to Nvidia’s advanced computing infrastructure, Arc researchers can carry out more complex projects, analyse larger datasets and reach results faster. The institute’s main research areas include cancer, immune system disorders and neurodegenerative diseases.
To develop Evo 2, Nvidia provided 2,000 H100 GPUs through the Nvidia DGX Cloud platform, which runs on the AWS cloud. DGX Cloud gives researchers short-term access to high-performance computing clusters, fostering innovation and scientific progress.