Digital biology and generative AI are revolutionizing drug development, surgery, medical imaging, and wearable technology—possibly the most significant shift to the healthcare sector since the invention of computers.
For more than ten years, NVIDIA has been laying the groundwork for this moment by developing deep domain experience, developing the NVIDIA Clara computing platform specifically for the healthcare industry, and growing its collaboration with a wide range of partners. Each year, NVIDIA GPU computing is already used by healthcare partners and customers for well over a billion dollars, both directly and indirectly through cloud partners.
R&D teams may now depict pharmaceuticals within a computer, marking a turning point in the $250 billion area of drug discovery.
Using newly developed generative AI techniques, drug development teams are able to produce or design unique compounds that are likely to have desired features by observing the fundamental building blocks of chemical sequence, structure, function, and meaning. These tools enable researchers to narrow down the pool of potential drugs to study, which eliminates the need for costly, time-consuming in-person testing.
NVIDIA BioNeMo, a generative AI platform that offers services to create, modify, and implement foundation models for drug discovery, is quickening this change.
Utilized by techbio, software, and pharmaceutical firms, BioNeMo provides a novel class of computational techniques for drug discovery and development, allowing researchers to incorporate generative AI to minimize and, in certain situations, completely eliminate studies.
NVIDIA has enhanced the computer-aided drug discovery ecosystem by investing in cutting-edge techbio companies like biopharmaceutical company Recursion, which is providing one of its foundation models for BioNeMo users, and biotech company Terray Therapeutics, which is using BioNeMo for AI model development. This is in addition to NVIDIA’s role in developing, optimizing, and hosting AI models through BioNeMo.
BioNeMo Boosts AI-Accelerated Drug Discovery with Precision
Pretrained biomolecular AI models for molecular optimization, protein structure prediction, generative chemistry, docking prediction, and other applications are becoming available through BioNeMo. Additionally, it makes models from computer-aided drug discovery businesses accessible to a wider audience via user-friendly APIs for customization and inference.
BioNeMo cloud APIs, which will be released in beta this month, are being integrated by drug discovery software companies, techbios, and large pharmas into platforms that provide computer-aided drug discovery workflows. BioNeMo teams use proprietary data to create or modify generative AI models.
The cloud APIs will now feature foundation models from three different sources: proprietary models created by NVIDIA partners, like Recursion’s Phenom-Beta for embedding cellular microscopy images, open-source models pioneered by international research teams and curated and optimized by NVIDIA, like the OpenFold protein prediction AI, and models created by NVIDIA itself, like the MolMIM generative chemistry model for small molecule generation.
MolMIM finds new molecules with the needed features and adheres to user-specified limitations, generating small molecules while providing users with more precise control over the AI production process. Researchers may instruct the model, for instance, to produce molecules with the same structures and characteristics as a certain reference molecule.
Pharma’s Phenomenal AI: Recursion Introduces Phenom-Beta Model to BioNeMo
The first hosting company to provide an AI model via BioNeMo cloud APIs is Recursion: Phenom-Beta is a vision transformer model that analyzes cellular microscope images and extracts information with biological significance.
This capacity can teach researchers about the behavior of cells and how they react to genetic engineering or potential drugs.
Using Recursion’s BioHive-1 supercomputer—which is built on the NVIDIA DGX SuperPOD standard architecture—Phenom-Beta was trained using the company’s publicly accessible RxRx3 dataset of biological pictures.
Recursion is adding more than 500 NVIDIA H100 Tensor Core GPUs to its supercomputer in order to advance the development of its foundation model. By doing this, its processing capability will increase fourfold, making it the most potent supercomputer that any biopharma company is likely to own and run.
The Way Businesses Are Using NVIDIA BioNeMo
NVIDIA BioNeMo is being used by an increasing number of scientists, biotech and pharma companies, and distributors of AI software to help research in biology, chemistry, and genomics.
Leading biotech company Terray Therapeutics is developing a generalized, multi-target structural binding model and is including BioNeMo cloud APIs. Additionally, the business trains chemistry foundation models with NVIDIA DGX Cloud to enable generative AI for small molecule design.
BioNeMo is being integrated into computational drug discovery applications by Innophore and Insilico Medicine, two businesses that specialize in protein engineering and molecular design. Protein design and drug discovery platform Catalophore from Innophore is incorporating BioNeMo cloud APIs. Furthermore, Insilico has included BioNeMo into its generative AI pipeline for early drug development. Insilico is a top participant in the NVIDIA Inception program for startups.
BioNeMo cloud APIs are being used by systems integrator Deloitte and biotech software business OneAngstrom to create AI solutions for their customers.
Academics, biotechs, and pharmaceutical companies use OneAngstrom’s SAMSON platform for molecular design, which now includes BioNeMo cloud APIs. By combining the Quartz Atlas AI platform with BioNeMo on NVIDIA DGX Cloud, Deloitte is revolutionizing scientific research. This combination opens up a new era of rapid drug discovery for biopharma researchers by providing them with state-of-the-art generative AI and unmatched data connection.
