Protein-protein etkileşimlerinin korunmuş bölge benzerliğine göre öbeklenmesi

iiiABSTRACTCLUSTERING PROTEIN-PROTEIN INTERACTIONSBASED ON CONSERVED DOMAIN SIMILARITIESAslı AyazM.S. in Computer EngineeringSupervisor: Assist. Prof. Dr. Uğur DoğrusözAugust, 2004Protein interactions govern most cellular processes, including signal transduction,transcriptional regulation and metabolism. Saccharomyces ceravisae is estimated to have16,000 protein interactions. Appereantly only a small number of these interactions wereformed ab initio (invention), rest of them were formed through gene duplications andexon shuffling (birth). Domains form functional units of a protein and are responsible formost of the interaction births, since they can be recombined and rearranged much moreeasily compared to innovation. Therefore groups of functionally similar, homologousinteractions that evolved through births are expected to have a certain domain signature.Several high throughput techniques can detect interacting protein pairs, resulting in arapidly growing corpus of protein interactions. Although there are several efforts forcomputationally integrating this data with literature and other high throughput data suchas gene expression, annotation of this corpus is inadaquate for deriving interactionmechanism and outcome. Finding interaction homologies would allow us to annotate anunannotated interaction based on already annotated known interactions, or predict newones. In this study we propose a probabilistic model for assigning interactions tohomologous groups, according to their conserved domain similarities. Based on thismodel we have developed and implemented an Expectation-Maximization algorithm forfinding the most likely grouping of an interaction set. We tested our algorithm withsynthetic and real data, and showed that our initial results are very promising. Finally wepropose several directions to improve this work.