Scripps receives 14 million-plus compute hours for AI antibody and protein research

Scripps Research has received more than 14 million hours of computing time to help scientists model how antibodies evolve and how human proteins move. Announced October 19, 2026, the Jen-Hsun and Lori Huang Foundation grant backs research aimed at choosing better vaccine candidates, finding therapeutic antibodies and exposing hidden targets for drugs.

The award supplies computing access rather than cash. It comes through the foundation’s partnership with CoreWeave, using NVIDIA graphics processing units, or GPUs. That computing capacity will support several Scripps projects, including an AI model of antibody development and simulations of proteins that have proved difficult to study.

Scripps describes the shared goal as moving AI beyond promising demonstrations toward reliable tools that scientists can use widely. Its researchers intend to use the additional computing power to train models on large datasets generated at the institute, with applications spanning vaccines, antibody therapies and small-molecule drugs.

Rehearsing an antibody response

Professor Bryan Briney’s Antibodies for All project aims to build what the team calls a Virtual Germinal Center. The AI system would model three connected parts of an immune response: how antibodies mutate, how they bind to their targets and how they are selected as the response develops.

The planned system will draw on millions of paired antibody sequences and detailed structural data. Researchers would use it to test vaccine candidates computationally and search for therapeutic antibodies, then send the most promising options into experimental testing. The model is intended to guide which candidates deserve that next step.

Scripps says this approach could reduce years of trial and error and improve responses to emerging pathogens and diseases that remain hard to treat. Those are intended research benefits, not outcomes demonstrated by the compute award. The project’s proposed workflow still includes experimental testing after the computational search.

Paired with Scripps’ world-class scientists and unique combination of foundational and translational research, we aim to shorten the journey from scientific breakthrough to something that can make a real difference for patients.

Andrew Ward, Scripps Research professor helping lead the foundation-supported research

Finding drug targets as proteins move

Professor Stefano Forli’s team is tackling a different problem: the “dark” proteome. Scripps uses that term for thousands of human proteins whose changing behavior has resisted structural study and kept them beyond the reach of drug development. The team plans to simulate those proteins at scale under realistic physiological conditions.

Rather than focus only on a protein’s structure, the simulations aim to capture how it flexes, folds and shifts. Those movements can briefly expose pockets that drugs might target. Forli’s system is intended to identify such openings as they appear, creating possible starting points for medicines against targets previously considered inaccessible.

Experimental inputs, then shared tools

Both teams will also use experimental data from the laboratories of Andrew Ward, Christopher Parker and Benjamin Cravatt. The grant therefore supports work combining computational models with measurements from several Scripps labs, rather than treating computing power as a substitute for biological data.

Jensen Huang, the foundation’s co-president, said the resulting tools would be openly available to scientists searching for new medicines. That commitment extends the intended audience beyond the teams receiving the computing allocation. Open availability is a plan for the research outputs, not a release of finished tools alongside this grant announcement.