Nvidia strengthens collaboration with pharmaceutical and biotech partners by hosting the first proprietary AI model for Phenomics.
Biotech company Amgen has long used Nvidia’s cloud services for its research. Now, the two companies have announced that Amgen’s deCODE Genetics division will install a new supercomputer called Freyja at its headquarters in Reykjavík, Iceland. Freyja will help create an atlas of human diversity for the discovery of drug targets and disease-specific biomarkers and develop AI-driven models for precision medicine.
The planned DGX SuperPOD will consist of 31 DGX H100 nodes and a total of 248 H100 Tensor Core GPUs. It will be able to train state-of-the-art AI models in days rather than months, enabling researchers to develop more efficient analysis and learning methods for health and therapeutic insights.
Since 1996, deCODE has collected more than 200 petabytes of anonymized human data from nearly three million individuals. This unique data can provide valuable insights into disease across population differences.
Nvidia’s BioNeMO hosts first proprietary AI model Phenom beta
Nvidia also offers BioNeMO, a generative AI platform specifically focused on drug discovery and development. Serving a $250 billion market, the platform enables research teams to virtually design drugs using more than a dozen generative AI models and cloud services, reducing the number of experiments required. They can analyze the basic molecular building blocks for sequence, structure, function, and significance, and design new molecules with the desired properties.
BioNeMO now offers basic models from three sources: Nvidia-trained models, open-source models contributed by global research teams, and proprietary models developed by Nvidia partners. The first partner is biopharmaceutical company Recursion, which is now making its Phenom beta model available to BioNeMo users for embedding cell microscopy images. Phenomics is an emerging transdisciplinary field that deals with the systematic study of phenotypes on a genome-wide scale.
Phenom-Beta is a vision transformer model that extracts biologically relevant features from cell microscopy images. According to the company, this capability can provide researchers with insights into cell function and help them understand how cells respond to drug candidates or genetic manipulations.
Nvidia itself now offers MolMIM. The model generates small molecules and is designed to give users more control over the AI generation process by identifying new molecules with the desired properties that meet the constraints set by the user. For example, researchers can instruct the model to generate molecules with structures and properties similar to a specific reference molecule. This limits the search space and is also intended to support applications such as drug discovery.