NVIDIA BioNeMo Now Offers a Huge Biomolecular Sciences Foundation Model.
Evo 2 Introduction
Evo 2, a potent new AI model developed with NVIDIA DGX Cloud on Amazon Web Services (AWS), offers information on proteins, RNA, and DNA in a variety of species.
Evo 2, a potent new foundation model that comprehends the genetic code for every area of life, is now available to scientists worldwide. It was developed on the NVIDIA DGX Cloud platform in partnership with Stanford University and the nonprofit biomedical research organisation Arc Institute, and it was unveiled today as the largest publicly accessible AI model for genomic data.
On the NVIDIA BioNeMo platform, Evo 2 is accessible to developers worldwide, including as an NVIDIA NIM microservice enabling simple, safe AI implementation.
The building blocks of DNA and RNA, Evo 2, were trained on a massive dataset of almost 9 trillion nucleotides. It can be used for biomolecular research applications such as predicting the shape and function of proteins based on their genetic sequence, finding new molecules for industrial and medical uses, and assessing the effects of gene mutations on their functionality.
Patrick Hsu, a core investigator and cofounder of the Arc Institute and an assistant professor of bioengineering at the University of California, Berkeley, stated that “Evo 2 is a significant milestone for generative genomics.” “It’s can pursue solutions in environmental science and healthcare that are unthinkable today by deepening the comprehension of these basic building blocks of life.”
With options to modify model parameters, users can create a range of biological sequences using the NVIDIA NIM microservice for Evo 2. Through the open-source NVIDIA BioNeMo Framework, a suite of accelerated computing tools for biomolecular research, developers can download the model and refine Evo 2 on their private datasets.
According to Brian Hie, an Arc Institute innovation investigator, the Dieter Schwarz Foundation Stanford Data Science Faculty Fellow, and an assistant professor of chemical engineering at Stanford University, “creating new biology has historically been a time-consuming, uncertain, and artisanal process.” “With Evo 2, will enable researchers to more easily design biological systems, allowing for the development of novel and useful advancements in a fraction of the time that would have been required in the past.”
Enabling Complex Scientific Research
The Arc Institute, which was founded in 2021 with $650 million from its original benefactors, gives academics multiyear support so they may concentrate on creative research rather than writing grants, enabling them to address long-term scientific concerns.
For eight-year, renewable terms that can be held in conjunction with faculty appointments at one of the institute’s university partners Stanford University, the University of California, Berkeley, or the University of California, San Francisco its core investigators are granted funding and access to cutting-edge lab space.
Researchers at the Arc Institute may work on more challenging topics, examine more datasets, and produce results faster by fusing this special research environment with NVIDIA’s accelerated processing capabilities and expertise. The diseases that its scientists are interested in include cancer, immunological dysfunction, and neurodegeneration.
By providing scientists with access to 2,000 NVIDIA H100 GPUs through NVIDIA DGX Cloud on AWS, NVIDIA expedited the Evo 2 project. Large compute clusters are temporarily accessible through DGX Cloud, allowing researchers the freedom to experiment. NVIDIA BioNeMo, a fully managed AI platform, comes with optimised software in the form of NVIDIA BioNeMo Blueprints and NVIDIA NIM microservices.
Additionally, NVIDIA engineers and researchers worked closely together on AI optimization and scaling.
Applications Across Biomolecular Sciences
Evo 2 can shed light on proteins, RNA, and DNA. The model, which was trained on a diverse range of species from many life domains, including bacteria, plants, and animals, can be used in scientific domains like materials science, healthcare, and agricultural biotechnology.
Long genetic information sequences, up to one million tokens, can be processed by Evo 2’s innovative model design. The connection between remote portions of an organism’s genetic code and the mechanisms underlying cell function, gene expression, and disease may become clearer to scientists due to this expanded understanding of the genome.
An AI model must process the largest possible portion of a genetic sequence at once in order to analyze how such complex biological systems work, as a single human gene contains thousands of nucleotides.
In the fields of medicine and drug development, Evo 2 may aid scientists in identifying the gene variations linked to a particular illness and creating new compounds that specifically target those regions to cure it. For instance, scientists from the Arc Institute and Stanford discovered that Evo 2 could 90% accurately predict whether previously unknown mutations would impact gene function in experiments using the BRCA1 gene, which is linked to breast cancer.
By offering insights into plant biology and assisting scientists in creating crop types that are more nutrient-dense or climate-resilient, the model could help address the world’s food shortages in agriculture. Additionally, Evo 2 could be used in other scientific domains to build proteins that degrade plastic or oil or to create biofuels.
It is comparable to sending a potent new telescope to the furthest reaches of the cosmos to deploy a model such as Evo 2. “It’s don’t yet know what to expect to find, but it’s clear there is a tonne of room for exploration.”
