Generative multiscale models of biomolecular conformational dynamics: from fluctuations to many-component assemblies - this research program aims to integrate physics-based modeling with artificial…

The National Institute of General Medical Sciences (NIGMS) awarded Stanford University a $410,256 Project Grant on September 15, 2025, under the Biomedical Research and Research Training program (CFDA 93.859) to develop generative multiscale models of biomolecular conformational dynamics. The primary deliverable is an integrated research program combining physics-based computational modeling with artificial intelligence techniques to advance conformational sampling and design of biomolecules. Over the five-year project period (through July 31, 2030), Stanford will develop two interconnected research products: (1) computational tools that leverage Monte Carlo techniques, generative pre-trained transformers, and physics-based molecular dynamics simulations to improve conformational sampling and design of single molecules including intrinsically disordered proteins, mRNA, and small molecules; and (2) experimentally-informed minimal models of many-molecule assemblies using nonequilibrium statistical mechanics to characterize self-organization dynamics in biological systems away from equilibrium.

The research program will deliver open-source software tools and methodologies that function as a "statistical microscope" for providing molecular-level insights into heterogeneous biomolecular assemblies, particularly structures like stress granules that are currently inaccessible to traditional structural studies. Research activities will be conducted at Stanford's Palo Alto, California facility in close collaboration with experimental partners to validate predictions and refine computational models. The outputs are intended to enable systems-level molecular design applications for small molecules, peptides, and RNA while advancing fundamental understanding of biomolecular self-organization dynamics.

Generative multiscale models of biomolecular conformational dynamics: from fluctuations to many-component assemblies - this research program aims to integrate physics-based modeling with artificial intelligence to EN- hance conformational sampling and design of biomolecules, with a view towards understanding self- organization dynamics in biological systems away from equilibrium. Our laboratory has made significant progress in developing generative neural networks for protein conformational sampling and designing state-of-the-art algorithms for molecular discovery. Building on this foundation, the program will PUR- SUE two interconnected projects over the next five years: 1) creating computational tools that combine physics and AI to improve conformational sampling and design of single molecules like intrinsically DIS- ordered proteins, mrna, and small molecules and 2) modeling and characterizing self-organization dynamics of many-molecule assemblies using nonequilibrium statistical mechanics. The research will leverage advanced monte carlo techniques, generative pre-trained transformers, and physics-based molecular dynamics simulations to search chemical space, model conformational heterogeneity, and bridge spatial scales. A key focus will be developing experimentally-informed minimal models of COM- plex biomolecular assemblies like stress granules, and using generative machine learning to move seamlessly between spatial scales. The long-term vision is to build an effective statistical microscope that can provide detailed molec- ular insights into heterogeneous assemblies that are currently inaccessible to structural studies. This approach could enable design at the "systems" level by combining experimentally and computationally guided small molecule, peptide, and RNA design. The program will work closely with experimental COL- laborators to test predictions and develop new models. All tools developed will be made open-source to benefit the broader scientific community. Ultimately, this research aims to advance our understanding of biomolecular structure and dynamics across scales, from single molecules to mesoscale assemblies, with implications for drug discovery and the engineering of novel biomaterials.