Two-Dimensional Multi-Target Detection

Abstract

Single-particle reconstruction from cryo-electron microscopy (cryo-EM) has recently emerged as a primary high-resolution structural determination method for biological molecules with key advan- tages over its predecessors, X-ray crystallography and nuclear magnetic resonance spectroscopy. Present methodology involves imaging many particles suspended in solution; electron microscopes generate image observations of this solution called micrographs, which contain embedded projections of the subject molecules. The traditional computational pipeline relies on detecting and extract- ing the particle locations within the micrographs and then recovering the volume from isolated sub-images of the projections. However, in high noise regimes, particle detection is impossible and standard procedures fail. This is of particular interest for small molecules that induce a low signal-to- noise ratio in the micrograph. In response, researchers have sought to develop methods independent of the location of the molecules and directly reconstruct particles in this regime. This project aims to expand a previously proposed autocorrelation analysis framework to avoid intermediate detection, and recover a projection from micrographs containing many occurrences of this projection up to in-plane rotation.