Identification of interaction sites of G protein-coupled receptors using machine learning techniques

Download
2014
Şahin, Mehmet Emre
G protein-coupled receptors (GPCRs), which play a crucial role in a host of pathophysiological pathways, form the largest and most divergent receptor family. Typically, they transmit outer signals to the inner cell by interacting with G-proteins. The emerging concept of GPCR dimerization has unsettled the classical idea that GPCRs function as monomeric units. Prediction of the interface residues of GPCR dimers is a challenging topic. The method proposed in this thesis trains itself with known interfaces from the literature and makes predictions using both the sequence and threedimensional structural information about GPCRs. The predictions are assessed by comparison to known interfaces in the literature. Our results show that the predictions are consistent with real interactions; however, further biological validation is still needed. During the development of the method, a new database was published for the use of the community: IntGPCR, the database of interacting GPCRs. IntGPCR contains information about interacting GPCRs, where the contents are curated from the literature. Up-to-dateness and the wealth of its contents, containing 309 interacting GPCRs curated from 348 articles, make IntGPCR a valuable resource for GPCR researchers. The other proposed method is about the classification of the GPCRs, serving to the requirement of an efficient and rapid classification to group the receptors according to their functions. GPCRsort, a new classification tool for GPCRs using the structural features derived from their primary sequences is proposed. Comparison experiments with the current known GPCR classification techniques show that GPCRsort is able to rapidly (in the order of minutes) classify uncharacterized GPCRs with 97.3% accuracy whereas the best available technique’s accuracy is 90.7%.
Citation Formats
M. E. Şahin, “Identification of interaction sites of G protein-coupled receptors using machine learning techniques,” Ph.D. - Doctoral Program, 2014.