Solvent-Driven Assembly of Intrinsically Disordered Peptides: Integrating Protein-Language Models with Atomistic-to-Mesoscopic Simulation

About This Grant

PROJECT SUMMARY The project supports ongoing efforts in the Shea group geared at developing new computational methodologies and tools to study the liquid-liquid phase separation (LLPS) of intrinsically disordered proteins (IDPs). Biomolecular condensates formed by LLPS play a range of vital physiological roles in the body, but under aberrant conditions, they can transition into amyloid fibrils, a process linked to disease. The project will meld artificial intelligence protein-language models with a multiscale computational framework to accurately simulate the dilute and dense LLPS phases, characterize the role of water, co-solvents, and high pressure in modulating assembly, and identify new LLPS-prone sequences in the human proteome. The proposal consists of three research projects. Project 1 involves the development of a tightly integrated multiscale computational approach bridging the atomistic to mesoscopic time and length scales. The relative entropy approach will be used to generate chemically accurate protein and water coarse-grained models from atomistic simulations, which will be used as input for efficient field theoretic simulations. The latter will be used to generate phase diagrams for the LLPS of the microtubule-binding Tau protein and Elastin-Like Polypeptides (ELPs), with field theoretic outputs backmapped to generate atomistic, solvated condensate structures that can be directly compared to experiment. Project two involves the development of new high pressure Kirkwood-Buff force fields for the osmolyte trimethylamine N-oxide (TMAO) from experimental Kirkwood-Buff Integrals, and their application to the study of TMAO’s counteraction of high-pressure denaturation of ELP condensates. Project 3 involves developing new artificial intelligence protein language model tools to mine the IDRome – the 28k proteome of intrinsically disordered regions – for new LLPS-prone and co-condensating sequences. The research will lead to state-of- the-art computational tools that will be deposited in Github and made freely available to the broad scientific community, to new physical insights into osmolyte and pressure modulation of LLPS, and to the discovery of new LLPS-prone sequences. The research will inform on conditions that promote functional forms of LLPS as well as lay the foundation for rational therapies for condensate-linked diseases.