Improving Initial Model Construction in Single Particle Cryo-EM by Filtering Out Low Quality Projection Images

Abstract

An important problem in the single-particle 3D reconstruction by cryo-electron microscopy (cryo-EM) is to construct the initial model of a macromolecule from its 2D noisy projection images at unknown random orientations. The methods for initial model construction are often based on ‘‘Angular Reconstruction’’ that computes the directions of the projection images by establishing a coordinate system. However, it is difficult to obtain the projection angles of the projection images which have low signal-to-noise ratio. In this paper we propose a method to improve the initial model construction by filtering out low quality projection images. The projection angles are usually represented by Euler angles $\upalpha$, $\upbeta$ and ${\upgamma }$. It is found that the quality of a projection image can be evaluated in the process of estimating its Euler angle ${\upgamma}$. After the low quality projection images are removed, the rest of the projection images are used to construct the initial model in our method. Based on the synchronization method for initial model construction proposed by Yoel Shkolnisky, it is found that using the proposed filtering method can successfully improve the initial model construction. It is also found that filtering using Euler angle ${\upgamma}$ estimation is better than filtering using good common line estimation in the initial model construction.