Evolutionarily motivated computational methods for analysis of protein sequences.

Abstract

The main goal of this thesis work was the development and application of a novel site-class model of evolution. The novelty of this approach was the introduction of site-specific substitution models that account for both quantitative and qualitative differences in selective pressures acting on various parts of proteins. Work presented in this thesis has mainly explored the applicability of such models to analyze sequences from G-Protein Coupled Receptors (GPCRs) and neurotransmitter transporters and its performance against other computational tools of sequence analysis. Furthermore, this thesis work aimed to develop novel computational approaches to solve two biological questions of interest, namely the specificity of G-Protein coupling of GPCRs and lipid exposure at transmembrane segments of membrane proteins. The development of such approaches was mainly driven by evolutionary and biophysical considerations. This work proved that a site-class model of evolution can be used successfully as a sequence analysis tool, allowing natural incorporation of evolutionary information and site-specificity. Furthermore, it showed that this model was more successful than other computational tools for predicting functional sites in proteins, confirming the importance of accounting for differences in selective pressures among different locations and for evolutionary relations among proteins when carrying computational analysis of their sequences. The secondary studies contained in this work were related to the main theme of the thesis in that they involved the proteins of interest in one case and use of evolutionary considerations in the other. The former study involved the prediction of G-Protein specificity of GPCRs. Results of this and further studies indicated that a hidden Markov model, incorporating evolutionary information among proteins, was most successful in such prediction. The latter study involved the use of evolutionary perspectives for creating tools to predict lipid exposure of transmembrane locations in membrane proteins. Results of this analysis indicated the complexity of this problem and proved that the site-class model is superior in such prediction in case of alpha-helical membrane proteins with complex packing of transmembrane helices. This thesis work represents an important step to emphasize the importance of evolutionary considerations while developing any type of computational tool to analyze biological data.