Frontiers in measuring and simulating molecular structures and processes.

Yale University, New Haven CT

Investigators

Abstract

Abstract Understanding the structure of molecular complexes and the way molecular structures interact is crucial to the understanding of biological processes and the development of drugs. Researchers study these structures and processes using empirical measurement methods and computational tools for structure prediction and simulation of processes. In recent years, significant advancements have been made in identifying the most stable and common conformations using empirical methods and computational tools, including some new machine- learning approaches. However, there has been more limited success in characterizing flexible complexes and intermediate molecular states in biological processes. Our group is particularly interested in cryo-electron microscopy (cryo-EM), an imaging technology that has revolutionized structural biology in recent years, and molecular dynamics (MD), the workhorse of simulation of molecular processes. Cryo-EM holds a long-standing promise to allow researchers to map diverse collections of molecular states in biological processes, including intermediate states. This promise is only partially fulfilled. We are developing the methods to deliver on this promise. In recent years, we have introduced a novel framework for mapping the conformation landscape using cryo-EM; this framework has been implemented in many of the recent algorithms. We have also developed software and made various other contributions to theory and methods in the area. In addition, we are developing novel techniques that will accelerate the notoriously slow MD simulations and we are developing algorithms that combine the computational modeling power of MD with the empirical power of cryo-EM to achieve results that neither technology can obtain on its own. In our work, we seek to combine robust interpretable mathematical modeling and algorithms with new and emerging data-driven machine learning algorithms to develop practical, robust, and interpretable methods for solving scientific questions about the structure and function of molecular complexes. Our group is also interested in broader statistical and computational methodology. We will develop more statistically rigorous and canonical methods and software for some of the classic statistical questions that arise in biology and medical research. These methods would improve the reliability of medical research and evaluation of drugs due to their advantageous statistical properties and canonical nature, reducing the risk of manipulation.