Exploring deep phylogenies using protein structure : a dissertation submitted in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Biochemistry, Institute of Natural and Mathematical Sciences, Massey University, Auckland, New Zealand

Abstract

Recent times have seen an exponential growth in protein sequence and structure data. The most popular way of characterising newly determined protein sequences is to compare them to well characterised sequences and predict the function of novel sequences based on homology. This practice has been highly successful for a majority of proteins. However, these sequence based methods struggle with certain deeply diverging proteins and hence cannot always recover evolutionary histories. Another feature of proteins, namely their structures, has been shown to retain evolutionary signals over longer time scales compared to the respective sequences that encode them. The structure therefore presents an opportunity to uncover the evolutionary signal that otherwise escapes conventional sequence-based methods. Structural phylogenetics refers to the comparison of protein structures to extract evolutionary relationships. The area of structural phylogenetics has been around for a number of years and multiple approaches exist to delineate evolutionary relationships from protein structures. However, once the relationships have been recovered from protein structural data, no methods exist, at present, to verify the robustness of these relationships. Because of the nature of the structural data, conventional sequence-based methods, e.g. bootstrapping, cannot be applied. This work introduces the first ever use of a molecular dynamics (MD)-based bootstrap method, which can add a measure of significance to the relationships inferred from the structure-based analysis. This work begins in Chapter 2 by thoroughly investigating the use of a protein structural comparison metric Qscore, which has previously been used to generate structural phylogenies, and highlights its strengths and weaknesses. The mechanistic exploration of the structural comparison metric reveals a size difference limit of no more than 5-10% in the sizes of protein structures being compared for accurate phylogenetic inference to be made. Chapter 2 also explores the MD-based bootstrap method to offer an interpretation of the significance values recovered. Two protein structural datasets, one relatively more conserved at the sequence level than the other and with different levels of structural conservation are used as controls to simplify the interpretation of the statistics recovered from the MD-based bootstrap method. Chapter 3 then sees the application of the Qscore metric to the aminoacyl-tRNA synthetases. The aminoacyl-tRNA synthetases are believed to have been present at the dawn of life, making them one of the most ancient protein families. Due to the important functional role they play, these proteins are conserved at both sequence and structural levels and well-characterised using both sequence and structure-based comparative methods. This family therefore offered inferences which could be informed with structural analysis using an automated method. Successful recovery of known relationships raised confidence in the ability of structural phylogenetic analysis based on Qscore to detect evolutionary signals. In Chapter 4, a structural phylogeny was created for a protein structural dataset presenting either the histone fold or its ancestral precursor. This structural dataset comprised of proteins that were significantly diverged at a sequence level, however shared a common structural motif. The structural phylogeny recovered the split between bacterial and non-bacterial proteins. Furthermore, TATA protein associated factors were found to have multiple points of origin. Moreover, some mismatch was found between the classifications of these proteins between SCOP and PFam, which also did not agree with the results from this work. Using the structural phylogeny a model outlining the evolution of these proteins was proposed. The structural phylogeny of the Ferritin-like superfamily has previously been generated using the Qscore metric and supported qualitatively. Chapter 5 recovers the structural phylogeny of the Ferritin-like superfamily and finds quantitative support for the inferred relationships from the first ever implementation of the MD-based bootstrap method. The use of the MD-based bootstrap method simultaneously allows for the resolution of polytomies in structural databases. Some limitations of the MD-based bootstrap method, highlighted in Chapter 2, are revisited in Chapter 5. This work indicates that evolutionary signals can be successfully extracted from protein structures for deeply diverging proteins and that the MD-based bootstrap method can be used to gauge the robustness of relationships inferred.