Manifold Bio Demonstrates Million-Scale Experimental Validation of AI-Driven Protein Binder Design with NVIDIA

BOSTON--(BUSINESS WIRE)--Manifold Bio, a platform therapeutics company pioneering direct-to-vivo drug discovery, today announced a joint study with NVIDIA to validate Proteina-Complexa, NVIDIA's generative model for protein binder design. Manifold used its proprietary platform to test 1 million binder designs against 127 targets in a single multiplexed experiment, measuring over 100 million protein-protein interactions. The study identified specific binders to 68% of targets tested, establishing Proteina-Complexa as a competitive generative model for protein binder design.

“Manifold Bio’s platform uniquely enabled this massively multiplexed study, which establishes Proteina-Complexa as competitive with state-of-the-art methods,” said Pierce Ogden, Ph.D., Co-founder and Chief Technology Officer of Manifold Bio. “We created our molecular synthesis and measurement technologies to unlock the full potential of generative AI design. This study provides a practical demonstration of scaling laws in de novo protein design, with more designs producing more hits, and makes the case for inference scaling when experimental throughput can keep pace.”

“Proteina-Complexa was built to generate protein binders at the speed and scale that drug discovery demands, powered by a novel architecture that redefines generative design,” said Anthony Costa, Director of Digital Biology at NVIDIA. “With test-time scaling, we’ve enabled the model to refine its logic before outputting a single sequence. What Manifold Bio brought to this collaboration is the rare ability to match experimental throughput to this high-fidelity output. Testing over a million designs against 127 targets in a single experiment validates that this scaled reasoning approach works, and demonstrates experimental throughput that matches the scale of generative AI.”

The experimental approach used in the study was first reported by Manifold in October 2025, when Manifold open-sourced mBER, an AI model for designing epitope-specific antibodies. Manifold used mBER to design and test more than 1.1 million VHH antibodies against 145 diverse targets, measuring over 100 million protein-protein interactions. Manifold’s core platform further integrates multiplexed in vivo studies to provide an additional layer of physiologically relevant data for its AI-designed binders.

The combination of de novo generative design and massively multiplexed measurement generated thousands of new annotated structures for future training, and sets a standard for matching the throughput of generative AI models with experimental validation at equivalent scale.

About Manifold Bio

Manifold Bio is a platform therapeutics company building the first AI-guided direct-to-vivo discovery platform. The company’s mDesign engine integrates AI-guided protein design with multiplexed in vivo data generation, creating a virtuous cycle of design and translation, building towards the virtual organism. In addition to advancing its own internal pipeline of novel therapeutics, Manifold collaborates with leading global pharmaceutical partners to accelerate the translation of organism-scale biology into meaningful clinical impact.

Learn more at https://www.manifold.bio/