Elucidating Allosteric Regulation of Spastin through Machine Learning

Abstract

Project ID: 170

Awards: Excellence in Research Mentoring; Excellence in Research Communication

Student Major: Chemistry-ACS

Project Advisor: Ruxandra Dima

Abstract: Allostery broadly refers to the ability of a protein to undergo structural changes in response to ligand binding at a location other than its active site. This phenomenon regulates protein function and thus motivates the development of computational methods that can be influential in providing detailed mapping of allosteric sites. Microtubules (MT) are found in eukaryotic cells and have critical roles in various cellular functions. MT severing enzymes, such as spastin, generate internal breaks in MTs that allow for these cellular events to take place, and have been shown to cause neurodegenerative disorders when mutated. In the active severing form, spastin forms a hexamer in the presence of the ATP and tubulin carboxy-terminal tails (CTTs), but this assembly has yet to be fully understood. Here, we used molecular dynamics simulations of a spastin monomer in order to identify the allosteric responses due to ligand binding. We utilized Markov State Models to find that the allosteric mechanism of the monomer is dependent on the ligand present. Using machine learning classification approaches, we identified allosteric regions that showed significant changes in various biochemically-relevant properties due to ligand binding that matched with experimentally-known allosteric sites. Finally, we identified the monomer's preference towards initial binding of ATP and the desire for the monomer to form monomer-monomer interfaces when ligands were present.