Speaker:  Dr. Wai Shing Tang

Host: Prof. JC Gumbart

Title: Bayesian Frameworks for Understanding Conformational Heterogeneity in Cryo-EM Data.

Abstract:

In the past decades, technological advancements in experimental and computational approaches, such as nuclear magnetic resonance (NMR) spectroscopy, cryo-electron microscopy (cryo-EM), machine learning (ML), and molecular dynamics (MD) simulations, have revolutionised structural biology. They provide comprehensive information into the structural dynamics of biomolecules, allowing scientists to unravel the complex relationships between molecular structural dynamics and biological function with unprecedented precision. This talk presents my Bayesian approach for extracting conformational heterogeneity from cryo-EM data. It uses the cryo-EM images to reweight a given conformational ensemble generated by an MD simulation. The reweighted ensemble approximates the biomolecule's free energy landscape, offering a physical interpretation of conformational heterogeneity in cryo-EM data. Nonetheless, this approach requires evaluating the image-to-structure likelihoods for thousands of cryo-EM images to dozens of structures, which can be computationally demanding. To extend the scalability and applicability of the Bayesian approach, I developed CryoLike, a Python package with GPU acceleration for evaluating the image-to-structure likelihood efficiently and robustly. Lastly, I present an application of the developed tools in extracting the free energy landscape from an experimental cryo-EM dataset of a flexible biomolecule.