NVIDIA Develops Evo 2

February 21, 2025

Nvidia has developed Evo 2, the largest artificial intelligence system dedicated to biological research. Announced in 2025, this system aims to revolutionise medicine and genetics. Evo 2 can read and design genetic codes across all forms of life. It uses nearly 9 trillion pieces of genetic information from over 128,000 organisms. This technology promises to transform research by identifying patterns in massive datasets that would take years to analyse manually.

About Evo 2

Evo 2 is a collaborative effort between Nvidia, the Arc Institute, and Stanford University.
It is built on Nvidia’s DGX Cloud platform, utilising 2,000 Nvidia H100 processors.
The model has 40 billion parameters, making it comparable to current large language models.
Evo 2 can process DNA sequences of up to 1 million nucleotides, allowing it to understand complex genetic relationships.

Applications in Medicine

Evo 2 has shown promise in identifying harmful mutations in genes such as BRCA1, which is linked to breast cancer.
In early tests, it achieved over 90% accuracy in predicting disease-causing mutations.
This capability could lead to more precise treatments, including targeted gene therapies.
Understanding gene variants is crucial for developing effective therapeutics.

Impact on Agriculture and Environmental Science

Beyond medicine, Evo 2 has potential applications in agriculture. Researchers believe it can help create crops that are more resilient to climate change. The technology may also assist in developing methods to break down environmental pollutants. This dual focus on health and environmental sustainability puts stress on the system’s broad applicability.

Advancements in Biological Design

Evo 2 represents leap in biological design. Previous models focused primarily on molecular levels. Evo 2 allows for the design of entire genomes. It has demonstrated the ability to construct mitochondrial genomes and eukaryotic chromosomes. This capability opens new avenues for research in various biological contexts.

Open Source and Community Engagement

Nvidia has made Evo 2 publicly available through the BioNeMo platform. The training data and model weights are open-source. This initiative aims to empower researchers worldwide to utilise Evo 2 for their projects. The collaborative approach encourages innovation and exploration in biological research.

Future Prospects

The Arc Institute plans to explore the concept of a virtual cell. This involves integrating genetic data with environmental information. The goal is to answer fundamental questions about diseases and their causes. Evo 2 serves as a foundational model for generative functional genomics, enabling researchers to build upon its capabilities.

« Previous Post Next Post »