We present Proteina-Complexa, a novel fully atomistic protein binder design framework unifying conditional generative modeling and optimization. This page is partitioned into two sections: Core model development and wet lab validation. We carried out an experimental design campaign with over 1 million binder candidates for 133 targets in collaboration with Manifold Bio, Viva Biotech, Novo Nordisk, Duke University, Cambridge University and LMU Munich

Really great collaboration between @ManifoldBio and @nvidia to test 1M de novo designed binders against 127 targets, measuring over 100 million potential protein-protein interactions! This was a great collaboration with some very exciting results. NVIDIA's new Proteina-Complexa method is SOTA for de novo minibinder design.

We're moving from a screening-driven to a generative design-driven paradigm in computational biology.
Scaling Data & Compute: Complexa trains on the Teddymer dataset—a vast library of synthetic domain-domain interactions. This allows the model to deeply learn the fundamental logic and physics of protein interactions.
The "o1 moment" for binders: Complexa doesn't just output a shape. It trades inference-time compute for accuracy, actively searching its learned structural intuition for the most physically plausible solution. Fully transformer-based, Complexa delivers ready-to-test binder sequences for proteins, small molecules, and even functional enzyme design.