Classification of Heterogeneous Particles in Cryo-EM Reconstructions

CVC Seminar Wed Sept 21 1:15-2:30pm ACES 4.304

Speaker: E. Greg Daniel

Title: Classification of Heterogeneous Particles in Cryo-EM Reconstructions

Abstract: One of the common approaches in single particle reconstruction of viruses is Cryo-Electron Microscopy. In this approach, a large sample of particles (with unknown orientations) are produced, and then an electron micrograph is created capturing the projections of all these particles. These 2d projections are then used to produce a 3d reconstruction of the virus particle. In current approaches, it is commonly assumed that all particles within the sample are completely homogeneous. However, in reality this assumption is often not the case. There are two common types of heterogeneities that appear in samples: structural (caused by slight mutations amongst particles) and conformational (due to the flexibility of particles). By not taking these heterogeneities into account, the resulting reconstructions often contain flaws. A solution to this problem is to classify heterogeneous particles into separate groups and then doing the reconstruction separately on each group. Thus, we are looking at statistical and machine learning approaches to create a classifier that will enable us to accurately separate the 2d projections of heterogeneous particles.

Comments are closed.