Cryo–electron microscopy (cryo-EM) is an experimental technique that produces high-resolution images of proteins. Using these images to recover a single 3D protein structure has been extremely successful. However, proteins are inherently flexible--the arrangement of atoms in 3D space, called a molecular conformation, is constantly changing. Recovering the entire probability distributions of conformations from cryo-EM data is a challenging inverse problem due to extreme noise levels and unknown imaging parameters. In this talk, I analyze ensemble reweighting, a recently introduced Bayesian approach that adjusts probabilities of conformations generated from molecular simulations to match experimental cryo-EM images. Using sensitivity analysis on the non-parametric maximum likelihood estimation problem, I identify when uninformative, noisy images can be discarded without losing information about rare conformations, thereby reducing the computational cost.

For more information, please visit our website: Math Frontier Seminar Website.