Font Size: a A A

A set of computational tools for the analysis and prediction of protein structures

Posted on:2005-02-25Degree:Ph.DType:Dissertation
University:Washington University in St. LouisCandidate:Welsh, Eric AFull Text:PDF
GTID:1450390011452696Subject:Biophysics
Abstract/Summary:
A set of computational tools has been developed to aid in the analysis and prediction of protein structures. These tools consist of a scoring function for the selection of native and near-native folds, a novel method of secondary-structure assignment, a method for the reconstruction of the all-atom protein backbone from a set of alpha carbon atoms, and a method for the structural superposition and alignment of two protein structures. The ProVal scoring function is used to select native and near-native structures from a set of candidate protein folds. In tests performed on a variety of decoy sets, the native (crystal) structure was identified 64% of the time. In cases where the native structure was not ranked first, it ranked in the top 10%. A new method of secondary-structure assignment has been developed that is tolerant of non-ideal secondary-structure geometries and agrees well with manual assignment. As there is no standard to compare to, comparisons of assignments are made to other widely used methods, the differences noted and explained. The all-atom backbone reconstruction method offers a useful compromise between fitting the input alpha carbon coordinates and correcting for random noise in the data. In tests performed on crystal structure coordinates, the method is able to accurately reconstruct 99% of residues involved in secondary-structure, with an accuracy over all residues of 96%. The addition of 1 A RMSD noise to the input alpha carbon coordinates has only a small effect on accuracy, correctly reconstructing 97% of residues involved in secondary-structure and 96% overall. The pair-wise structure superposition method is more robust and generates better superpositions than all other methods tested. It is also able to correctly superimpose and align structures with permuted secondary-structure order. Multiple pair-wise superpositions can be used to generate structure based multiple sequence alignments. Families of structures with sequence homology too low to be aligned with traditional sequence based methods may be aligned using this method.
Keywords/Search Tags:Structures, Protein, Tools, Method
Related items