Computational chemistry in protein research: Structural models of cytochrome oxidase, alpha-amylase and antibody-antigen complexes: Dissertation

A collection of computer tools designed for amino acid sequence analysis, model building and molecular simulation of proteins is reviewed, with emphasis on case studies of three proteins for which X ray structures were not available but sequence and mutant data warranted an investigation by computer aided molecular modelling.A program, MaxSprout, was written for building a full set of protein coordinates given only a Ca trace, The algorithm optimized backbone stereochemistry by combining fragments taken from a database of known structures, taking care that they overlap smoothly. Sidechains were added from a rotamer library, and optimized for sterical fit. The results obtained with this simple algorithm compare favourably with (a) manual and (b) time consuming computerized search methods.Structural models of the redox centres in cvtochrome oxidase were based on a synthesis of evolutionary conservation, predicted membrane topography and known spectroscopic properties.The redox centres in subunit I were modelled with four helical segments treated as rigid blocks of hard spheres.Subunit 1I was proposed to contain a copper binding domain homologous to that of the blue copper proteins.The structural models are in agreement with many experimental observations.A structural model of Bacillus stearothermophilus a amylase was built based on a multiple sequence alignment of a amylases.There was 24 % sequence identity to Taka amylase A, which was used as a 3 D template.The changes in the activity of ninety eight mutant enzymes could be rationalized on the basis of their location in the 3 D model.The coherence of model and mutant data support the structural ideas.Oxazolone binding affinities of a series of natural anti oxazolone antibodies with known sequences but unknown structures were used to test the performance of theoretical methods for future protein design purposes.A strongly homologous template was available to model the protein, and the ligand was docked using a combination of interactive graphics and molecular dynamics.The comparison of modelled complexes between the ligand and idiotypic or mature antibody suggest plausible explanations for the experimentally observed increase in affinity.