Protein-protein Interaction Network Alignment and Evolution

Network alignment is an emerging analysis method enabled by the rapid large-scale collection of protein-protein interaction data for many different species. As sequence alignment did for gene evolution, network alignment will hopefully provide new insights into network evolution and serve as a new bioinformatic tool for making biological inferences across species. Using new SH3 binding data from Saccharomyces cerevisiae, Caenorhabditis elegans, and Homo sapiens, I construct new interface-interaction networks and devise a new network alignment method for these networks. With appropriate parameterization, this method is highly successful at generating alignments that reflect known protein orthology information and contain high network topology overlap. However, close examination of the optimal parameterization reveals a heavy reliance on protein sequence similarity and fungibility of other data features, including network topology data, an observation that may also pertain to protein-protein interaction network alignment. Closer examination of interactomic data, along with established orthology data, reveals that protein-protein interaction conservation is quite low across multiple species, suggesting that the high network topology overlap achieved by contemporary network aligners is ill-advised if biological relevance of results is desired. Further consideration of gene duplication and protein binding sites reveal additional PPI evolution phenomena further reducing the network topology overlap expected in network alignments, casting doubt on the utility of network alignment metrics solely based on network topology. Instead, I suggest a new framework to think about protein-protein interaction network alignment focused on generating and validating small-scale inferences. I create a prototype alignment visualization and analysis tool to facilitate this approach, which will hopefully aid researchers in learning more about the mechanisms of network evolution and how network alignment can model them.Ph.D