The NVIDIA BioNeMo platform now provides AI for biomolecular sciences

Evo 2, a groundbreaking AI model capable of deciphering genetic codes across all life forms, has now been made accessible to scientists worldwide. Launched as the largest publicly available AI model for genomics, it was developed through a collaboration between the nonprofit Arc Institute and Stanford University, leveraging the NVIDIA DGX Cloud platform. This powerful model is now available to developers via NVIDIA’s BioNeMo platform, including as an NVIDIA NIM microservice, which ensures secure and efficient AI deployment.

Built on a vast dataset of nearly nine trillion nucleotides, Evo 2 can be applied in numerous biomolecular research areas. Its capabilities include predicting the structure and function of proteins from genetic sequences, identifying new molecules for medical and industrial use, and assessing the impact of genetic mutations. According to Patrick Hsu, Arc Institute cofounder and core investigator, this marks a significant leap for generative genomics, offering new possibilities in healthcare and environmental science.

With Evo 2, researchers can generate various biological sequences while adjusting model parameters through the NVIDIA NIM microservice. Those interested in customizing the model with their proprietary data can do so via the open-source NVIDIA BioNeMo Framework, which provides cutting-edge tools for biomolecular research. Brian Hie, assistant professor at Stanford University, emphasized that Evo 2 is transforming biological design, making it faster and more accessible for researchers to develop complex systems.

Established in 2021 with $650 million in funding, Arc Institute focuses on solving long-term scientific challenges by providing researchers with stable, multi-year funding. Its investigators receive top-tier lab space and resources, allowing them to concentrate on innovation rather than securing grants. Working alongside university partners like Stanford, UC Berkeley, and UC San Francisco, Arc researchers target diseases such as cancer, immune dysfunction, and neurodegeneration.

To accelerate Evo 2’s development, NVIDIA provided access to 2,000 H100 GPUs through DGX Cloud on AWS, a flexible and powerful AI research platform. This collaboration between Arc Institute and NVIDIA’s AI experts allowed the model to scale efficiently while processing massive datasets at an unprecedented pace.

Evo 2 is designed to unlock insights into DNA, RNA, and proteins across diverse species, spanning plants, animals, and bacteria. Its unique model architecture can analyze sequences up to one million tokens long, offering a broader view of the genome and deepening scientists’ understanding of cell function, gene expression, and disease mechanisms. Hsu highlighted that the model’s ability to process entire genetic sequences at once is key to advancing research in complex biological systems.

In medicine and drug discovery, Evo 2 could help identify gene variants linked to specific diseases and facilitate the design of precise molecular treatments. In recent tests involving the BRCA1 gene, which is associated with breast cancer, the model achieved a 90% accuracy rate in predicting whether previously unknown mutations would impact gene function.

Beyond healthcare, Evo 2 could revolutionize agriculture by enhancing plant biology research, aiding in the development of climate-resilient and nutrient-rich crops to address food security challenges. Additionally, it has potential applications in environmental science, such as engineering proteins to break down pollutants like oil and plastic.

Describing the impact of Evo 2, Arc’s chief technology officer Dave Burke likened it to launching a powerful new telescope into space-an instrument of immense potential that is set to reveal discoveries far beyond current expectations.