Computational methods for noisy high-dimensional images and application to Cryogenic Electron Microscopy (cryo-EM) workflow

Abstract

With the recent advancement of GPU-accelerated computations and algorithms, cryo-EM has become a mainstream technique to solve 3D structures of macro-molecules at near-atomic resolution. Remarkably, the first dynamic movie of 2019-nCoV Spike trimer structure is derived from cryo-EM within 12 days using computational methods. This further demonstrated that cryo-EM had been a powerful technique with high efficiency to provide crucial medical insight for developing vaccines or drugs. However, the data characteristics of cryo-EM images include strong noise, huge dimension, large sample size and high heterogeneity with unknown orientations have made analysis very challenging. In the literature, dimension reduction and clustering play an essential role in overcoming the challenges above. The traditional methods employed in the field, however, are not well suited for the scenario and they face bottleneck either in computation or performance. In this talk, I will first address the importance of cryo-EM image processing in structural biology. Second, I will discuss our proposed dimension reduction strategy called two-stage dimension reduction (2SDR) and clustering approach called DRMRA which alleviate the computation burden and improves performance over existing methods. Specifically, I will elaborate on how to utilize 2SDR to enhance the workflow and present a unified framework that enables us to select the best processing steps to deliver a reliable 3D initial model. Based on the initial model, I will describe how to perform 3D conformation analysis. Finally, future research directions in 3D conformation analysis will be given.