(A) novel rotation angle estimation algorithm for Cryo-EM single particle reconstruction

Abstract

In this study, whole system for Cryo-EM single particle reconstruction is designed. This system is mainly composed with two parts, projection/backprojection algorithm and estimation of particle orientations in 3D space. Direct use of conventional distance driven approach for Cryo-EM purpose will result in inaccurate approximation, which results in reconstruction artifacts. Distance Driven (DD) approach is successfully applied for projection/backprojection purpose in Cryo-EM imaging area after modification and improvements. The interpolation process in DD is done by calculating the overlap area between the detector and pixel boundaries. This procedure removes some interpolation artifacts and reduced the computational complexity. Estimation based on Polar Fourier Transform (PFT) is utilized for projection view determination. Due to the rotation invariant property of PFT, the reference library is effectively reduced, so estimation process can be speeded up. A global angle searching method is first applied for estimation. The intrinsic drawbacks of this approach result in large reference library, high computational complexity, artifacts due to non-uniform distribution of estimated projection views, and low resolution due to large sampling interval of reference library. To overcome these problems, angle estimation based on icosahedrons symmetries is proposed. With this method, reference library size is effectively reduced. Sampling interval of reference library can be as small as 1 degree, which make it possible for accurate estimation. Artifacts are removed since estimated projection views are uniformly distributed along 4 pi steradian. Experimental results of synthetic and real data are given and analyzed.