Covariation Methods in Proteomics

Covariation is a commonly observed process that can occur in biological systems where two characteristics experience change as influenced by their mutual feedback. This phenomenon occurs at varied biological scales, from macroscopic to molecular, and suggests an interaction between the evolving characteristics. Increasing availability of sequence data affords the material for examination of molecular covariation. This thesis examines molecular covariation using different elements of proteomic data and applies computational methods to infer protein functional pathway membership, protein-protein interactions and protein structure. We examine molecular covariation using stratified units of proteomic information - entire proteomes, comparison between (inter-) protein sequences and finally comparison within (intra-) protein sequences. Using entire proteomes, phylogenetic profiling assists in understanding the loss of the essential chaperonin GroEL in some species of Mycoplasma, and suggests a role for the chaperonin in adhesion-invasion. Diving further into the proteomic detail, we probe distances between protein sequences, using our MMM algorithm to detect covariation between phylogenetic trees as in indication of protein-protein interaction. In the Human co-evolution network, we find two subnetworks divided on the basis of evolutionary age. Finally, we introduce multi-dimensional mutual information (mdMI) as a means to detect intra-protein contact points through correlated mutation analysis and show that mdMI's accuracy is comparable to leading methods despite low computational time. Additionally, we also explore evidence that the physiochemical environment can unilaterally influence covariation score. The covariation methods applied and analyzed here represent a general approach to solving broad biological problems.Ph.D