Estimation of interatomic distance distribution of protein molecules from small angle scattering (SAS) images.

摘要

Small-angle scattering (SAS) is a frequently used technology to study the global structure of biological macromolecules in solution. The distribution of the molecules interatomic distances is known as the P(r) curve and represents the maximum structural information that can be deterministically derived from SAS data. It provides valuable information about the shape and orientation of the molecule and is used in many computational shape reconstruction programs. A SAS experiment generates a series of high resolution images. Traditionally, these images are summarized into a scattering curve and indirect transformation methods are used to reconstruct P(r). This data reduction procedure, however, utilizes some strong assumptions about the image data that can result in incorrect standard errors and possibly a biased P(r) estimate.;In this dissertation, we propose a new methodology that directly analyzes the complete scattering images. We build a multivariate spatial model for the scattering intensities of a molecule only image, thereby accounting for additional sources of uncertainty such as the beam center location of the image and spatial correlation among pixels. A Bayesian estimation method, utilizing Markov Chain Monte Carlo (MCMC), is employed. A posterior distribution of P(r) is obtained and can be used in these shape reconstruction programs. The performance of our method is demonstrated using both simulated and real experiments of protein samples in solution.