Nvidia collaborates with Genentech to accelerate AI-driven drug discovery

Nvidia is working with Roche Group’s Genentech to help the company pioneer the use of generative AI in drug discovery. The companies today announced a multi-year strategic collaboration that will see Genentech combine its own models and datasets with Nvidia’s AI computing stack to advance AI research — and expedite the discovery and delivery of novel therapies to people.  

The partnership, as the companies said in a blog post, will revolve around the use of Nvidia’s DGX Cloud and BioNemo offerings. Nvidia will share these offerings with Genentech’s scientists and help them optimize and scale their models and then use the learnings to further develop its product.

“We were the first biotech company to leverage molecular biology for drug discovery and development, which changed the world. We pioneered antibody therapeutics that became the paradigm of treatment. And now, we have brought AI, the lab and the clinic together to uncover otherwise inaccessible patterns in vast quantities of data and to design experiments to test those patterns. Collaborating with Nvidia, and introducing generative AI, has the power to turbocharge the discovery and design of therapeutics that will improve the lives of patients across the world,” Aviv Regev, executive vice president and head of Genentech Research & Early Development (gRED), said in a statement.

How Nvidia will help Genenetech with drug discovery

Drug discovery has long been a complex and time-consuming process, where teams have to research different molecules and their interactions to see what works to target a particular disease. Genentech has been exploring the potential of AI in this space, which can learn from large datasets to quickly identify potential drug molecules and interactions.

The organization’s AI/ML teams have already developed proprietary foundation models across numerous research areas, including diverse therapeutic modalities. Now, with this partnership with Nvidia, they aim to tap the DGX Cloud with BioNemo to further optimize these custom algorithms and accelerate the time to value in drug discovery.

Nvidia DGX Cloud will provide Genentech researchers with dedicated instances of AI supercomputing while BioNemo will give the ability to pre-train or fine-tune their models on those instances. The organization also plans to integrate BioNemo cloud APIs directly into its drug discovery workflows.

On the optimization side, Nvidia said its offerings will accelerate the training and inference of generative models used in Genentech’s “lab in the loop” framework, where extensive experimental data is fed into models for iterative development. When ready, the models uncover patterns and make new, testable predictions of molecular design, allowing scientists to assess those predictions in the lab and feed the results back to improve the underlying algorithm for the development of better therapies. 

For improving the custom models, Nvidia’s offerings will use Genentech’s extensive molecular and biological datasets. However, the organization did note that the company will not have direct access to this information unless explicitly granted for a particular project.

Over the next few years, Nvidia hopes to gain insights into AI-related challenges in drug discovery with the help of this partnership. These insights will ultimately inform the advancement of BioNemo and related products to better meet the needs of enterprises in the sector, the company said.