In an exciting advancement for the field of genomics, scientists around the globe now have access to Evo 2, a groundbreaking foundation model designed to decode the genetic language of all life forms. Announced today, this model stands as the largest publicly accessible AI model for genomic data, developed through a collaborative effort by the nonprofit biomedical research organization Arc Institute and Stanford University. Built on the NVIDIA DGX Cloud platform, Evo 2 is set to revolutionize biomolecular research.
Accessible to developers worldwide via the NVIDIA BioNeMo platform, Evo 2 is also offered as an NVIDIA NIM microservice, facilitating secure and straightforward AI deployment. This advanced model has been trained on a staggering dataset comprising nearly 9 trillion nucleotides—the fundamental units of DNA and RNA. Evo 2’s capabilities extend to predicting protein structures and functions based on genetic sequences, identifying new molecules for healthcare and industrial applications, and assessing the impact of genetic mutations on their functional properties.
Patrick Hsu, co-founder of the Arc Institute and a core investigator, as well as an assistant professor of bioengineering at the University of California, Berkeley, highlighted the significance of Evo 2. He remarked, “Evo 2 signifies a substantial leap forward for generative genomics. By deepening our understanding of the essential building blocks of life, we can pursue unprecedented solutions in healthcare and environmental science.”
The NVIDIA NIM microservice for Evo 2 empowers users to generate diverse biological sequences, with customizable settings to tweak model parameters. Developers interested in refining Evo 2 using their own datasets can download the model through the open-source NVIDIA BioNeMo Framework. This framework is a collection of accelerated computing tools tailored for biomolecular research.
Brian Hie, an assistant professor of chemical engineering at Stanford University and an Arc Institute innovation investigator, stated, “Designing new biological systems has traditionally been a painstaking and unpredictable task. Evo 2 democratizes the design of complex biological systems, enabling researchers to make significant advancements much more rapidly.”
### Enabling Complex Scientific Research
Founded in 2021 with $650 million from its founding donors, the Arc Institute is committed to addressing long-term scientific challenges. It achieves this by providing scientists with multiyear funding, allowing them to focus on innovation rather than being bogged down by the need for grant writing.
The institute’s core investigators receive cutting-edge lab facilities and eight-year, renewable funding terms. These terms can be held alongside faculty appointments at partner universities like Stanford University, the University of California, Berkeley, and the University of California, San Francisco.
This unique research environment, combined with NVIDIA’s expertise in accelerated computing, enables Arc Institute researchers to undertake more complex projects, analyze larger datasets, and achieve results more swiftly. The scientists focus on disease areas such as cancer, immune dysfunction, and neurodegeneration.
To expedite the Evo 2 project, NVIDIA provided scientists with access to 2,000 NVIDIA H100 GPUs through the NVIDIA DGX Cloud on AWS. DGX Cloud offers short-term access to large compute clusters, giving researchers the flexibility to innovate. The fully managed AI platform includes NVIDIA BioNeMo, which features optimized software like NVIDIA NIM microservices and NVIDIA BioNeMo Blueprints.
NVIDIA researchers and engineers played a crucial role in scaling and optimizing the AI processes.
### Applications Across Biomolecular Sciences
Evo 2 offers valuable insights into the realms of DNA, RNA, and proteins. Trained on a diverse range of species, including plants, animals, and bacteria, the model finds applications in fields like healthcare, agricultural biotechnology, and materials science.
Evo 2 employs an innovative model architecture capable of processing extensive sequences of genetic information, up to 1 million tokens in length. This broader view of the genome holds the potential to unlock scientists’ comprehension of the links between distant parts of an organism’s genetic code and the mechanics of cell function, gene expression, and disease.
Hsu explained, “A single human gene can contain thousands of nucleotides. For an AI model to effectively analyze such intricate biological systems, it must process the largest feasible segment of a genetic sequence simultaneously.”
In healthcare and drug discovery, Evo 2 could assist researchers in identifying which gene variants are linked to specific diseases and in designing novel molecules that precisely target those areas to treat the disease. For instance, in studies involving BRCA1, a gene associated with breast cancer, Evo 2 demonstrated a remarkable ability to predict with 90% accuracy whether previously unrecognized mutations would impact gene function.
In agriculture, Evo 2 could play a pivotal role in addressing global food shortages by providing insights into plant biology and aiding scientists in developing crop varieties that are more resilient to climate change or higher in nutritional value. Additionally, in other scientific disciplines, Evo 2 might be instrumental in designing biofuels or engineering proteins that can break down oil or plastic.
Dave Burke, Arc’s chief technology officer, likened the deployment of Evo 2 to sending a powerful new telescope to the farthest reaches of the universe. “We understand that there is enormous potential for exploration, but we have yet to discover what awaits us,” he noted.
For more information, you can visit NVIDIA’s official website.
This article offers an overview of the significant advancements brought about by Evo 2 in the field of genomics. As this model becomes accessible to more researchers, we can anticipate exciting developments across various scientific domains, shaping the future of healthcare, agriculture, and environmental science.
For more Information, Refer to this article.
