Roodmus Talk

This talk by Joel Greer (STFC) was recorded on 26 July 2024 at the CCPBioSim/CCP-EM talk.
Title: Using MD simulation-driven synthetic data to analyse heterogeneous single particle averaging methods in cryo-EM
Abstract: Single particle averaging (SPA) relies on computationally combining thousands of images in order to reconstruct a 3D representation of a biological macromolecule. As a result, these data sets are assumed to be representative of the different molecular conformations accessible under the experimental conditions prior to vitrification. Current standard practice is to employ classification and filtering methods which may allow a discrete number of conformational states to be reconstructed. Recently developed heterogeneous reconstruction algorithms (HRAs) to analyse continuous heterogeneity rely on machine learning methods employing low-dimensional latent space representations. The non-linear nature of many of these methods pose challenges to their validation and interpretations. We believe these methods would benefit from in-depth benchmarking using high quality synthetic data and concomitant ground truth information. Here we present Roodmus, a framework for the simulation and subsequent analysis of cryo-EM micrographs containing conformationally heterogeneous particles. Conformational heterogeneity is sourced from molecular dynamics simulations. We will demonstrate the simulation of such datasets and the analysis of the application of HRAs to them. Finally, we want to discuss envisioned future work to build synergy between heterogeneous reconstructions in cryo-EM and molecular dynamics by utilising simulated and experimental datasets.

[@StevenNess]

coool