Advances in protein structure determination and protein structural feature prediction using computer science.

摘要

Proteins are one of the basic building blocks of life. A thorough understanding of protein structure is fundamental all of biology and medicine. Central to studying proteins is the dogma that a protein's primary sequence gives rise to the protein's tertiary structure and a protein's tertiary structure determines how the protein functions in the cell.;In this thesis, several advances in both protein structure determination and protein structural feature prediction are studied. In Chapter 1, an introduction to protein structure is provided as well as a brief survey of methods for obtaining protein structures. In Chapter 2, a novel method is developed for increasing the size of proteins that can be determined using NMR spectroscopy. This method includes an objective function and set of constraints as well as a set of parameters estimated from NMR spectra databases and a novel dynamic programming algorithm for solving the optimization problem.;In Chapter 3, two integrated systems for predicting protein structural features are described. One system, SCRATCH, focuses on the prediction of various structural features of globular proteins. The other system, TMBpro, focuses on the prediction of the tertiary structure of beta-barrel membrane proteins.;In Chapters 4 and 5, new predictors of both continuous and discontinuous B-cell epitopes are described. The predictor of continuous B-cell epitopes, COBEpro, combines a support vector machine with the primary sequence and predicted secondary structure and relative solvent accessibility. The predictor of discontinuous B-cell epitopes, BEpro, combines various sequence and structural features to calculate an antigen propensity score. Several rigorous statistical tests are performed in Chapter 6 to assess the significance of COBEpro and BEpro.;The advances in this thesis have built on the success of prior efforts and have pushed the bounds of the study of protein structure. The predictors and systems are competitive with other published methods and have provided new insight in to the various problems studied.