Scientists worldwide now have access to Evo 2 a powerful base model that understands the genetic code across all domains of life This tool the largest publicly available AI model for genomic data was developed in a joint collaboration between the nonprofit Arc Institute Stanford University and on the NVIDIA DGX Cloud platform
Evo 2 is available to developers through the NVIDIA BioNeMo platform and in the form of NVIDIA NIM microservices for secure and easy deployment Trained on a massive dataset of 9 trillion nucleotides the building blocks of DNA and RNA it can be used in biomolecular research to predict protein structure identify new drug molecules and assess genetic mutations
One of the key highlights of Evo 2 is its novel architecture which enables the processing of long sequences of genetic information up to one million tokens This capability helps scientists better understand the connections between distant parts of an organism genetic code and cellular mechanisms and diseases For example researchers found that Evo 2 can predict the effect of unknown mutations in the BRCA1 gene related to breast cancer with 90 percent accuracy
In addition to medicine the model has applications in agriculture for developing climate resilient crops and in environmental science for designing plastic degrading proteins NVIDIA played a key role in accelerating this research project by providing access to 2 000 H100 GPUs and technical optimizations