دانلود رایگان مقاله BioEM : محاسبات GPU - شتاب گرفته استنتاج بیزی از تصاویر میکروسکوپ الکترونی

Abstract

In cryo-electron microscopy (EM), molecular structures are determined from large numbers of projection images of individual particles. To harness the full power of this single-molecule information, we use the Bayesian inference of EM (BioEM) formalism. By ranking structural models using posterior probabilities calculated for individual images, BioEM in principle addresses the challenge of working with highly dynamic or heterogeneous systems not easily handled in traditional EM reconstruction. However, the calculation of these posteriors for large numbers of particles and models is computationally demanding. Here we present highly parallelized, GPU-accelerated computer software that performs this task efficiently. Our flexible formulation employs CUDA, OpenMP, and MPI parallelization combined with both CPU and GPU computing. The resulting BioEM software scales nearly ideally both on pure CPU and on CPU+GPU architectures, thus enabling Bayesian analysis of tens of thousands of images in a reasonable time. The general mathematical framework and robust algorithms are not limited to cryo-electron microscopy but can be generalized for electron tomography and other imaging experiments.

5. Discussion

The BioEM method provides an alternative approach to structurally characterize biomolecules using electron microscopy images. By calculating the posterior probability of a model with respect to each individual image, it avoids information loss in averaging or classification, and allows us to compare structural models according to their posterior probability. Bayesian analysis methods, such as Relion [13,14], have been enormously successful in EM, contributing much to the resolution revolution [1]. However, the main use of Relion is to reconstruct 3D densities from projection images, and not to rank or compare existing structural models. It also differs from BioEM in the integration scheme and optimization algorithms (for a comparison see Supplementary Table S1). BioEM requires relatively few images to discriminate the correct model within a pool of plausible structures (e. g., <1500 particles for GroEL [34]), whereas, for full 3D reconstructions, Relion typically requires tens to hundreds of thousands of particles and costly computational resources to implement the multiple methods that select, classify, and polish the particles as well as refine the 3D maps. Beyond applications in studies of highly dynamic systems [34], we envision that BioEM can complement traditional 3D reconstruction techniques in the first steps of classification by assigning accurate orientations and single-particle PSF estimations, and in the last steps of refinement by validating the final 3D models. In addition, the BioEM method can be applied to problems where reconstruction techniques fail, e. g., when there are few particle images that acquire preferred orientations or when the system is flexible.