Company will extend its state-of-the-art PoET models to predict protein function through structural dynamics
OpenProtein.AI, a leader in AI-powered protein engineering software, today announced it has been selected as a performer on the Defense Advanced Research Projects Agency's (DARPA) Network of Optimal Dynamic Energy Signatures (NODES) program. The award, which began in March 2026, tasks OpenProtein.AI with developing a new generation of AI models capable of predicting protein function through structural dynamics.
Deep learning has transformed protein structure prediction. Yet, the function of >99% of proteins remains unknown. Current structure predictors cannot fill this gap. They predict frozen snapshots, but proteins are dynamic machines in constant motion, and it is that motion that determines their function. DARPA’s NODES program seeks to address this disparity. NODES is sponsoring the development of biophysics-inspired deep learning tools to predict biological function by identifying signatures of protein motion. The ability of these models to rapidly characterize novel or potentially engineered protein sequences could prove decisive in detecting and responding to biological threats.
OpenProtein.AI will pursue this goal by extending the capabilities of its PoET protein foundation models. Already PoET-2 features state-of-the-art sequence co-evolution capabilities and the ability to predict function from sequence. OpenProtein.AI’s objective under the NODES program is to extend these capabilities to better characterize protein complexes and predict structural dynamics. The extended models will learn from a dataset of 10,000 protein structural ensembles generated through enhanced molecular dynamics simulations, the largest of its kind. The goal is a generative framework capable of producing full structural ensembles in minutes with accurate thermodynamic weights, roughly 1,000 times faster than conventional simulation.
“Evolution has written the rules of protein motion into the sequences of natural proteins, but we have never had the data at scale to decode them,” said Tristan Bepler, Ph.D., CEO of OpenProtein.AI. “With this award, we will generate the data and build the next generation of foundation models required to decode protein dynamics and function. This will unlock new capabilities for predicting binding affinity, allosteric regulation and thermodynamic stability.”
These capabilities will directly advance OpenProtein.AI's commercial platform, which serves pharmaceutical and biotechnology companies applying AI to design-build-test cycles for antibodies, enzymes, and other proteins. The technology also has significant implications for national security, where the ability to rapidly predict protein function from sequence at genome scale could accelerate biological threat assessment and medical countermeasure development.
Robert Rankin
OpenProtein.AI
contact@openprotein.ai
