Study On The Parallel Prediction Of G-Protein Coupled Receptor Structure

G-protein couple receptor(GPCR) is used as a major drug target and plays animportant role in pharmacy because of its structure feature and its signifcant efect insignal transmission. It is very difcult to achieve GPCR’s structure by biochemistryexperimental method, so it’s signifcant to predict GPCR’s structure by computermethod. This thesis uses the parallel approach to study structure prediction problemsof GPCR’s backbone and side-chain.Considering the higher homology of the GPCR’s transmembrane helix region andthe lower homology of the loop region, this thesis proposes a parallel GPCR’s back-bone structure prediction method, pGPCR, which is based on current popular GPCR’sstructure prediction methods. This approach constructs the transmembrane region andloop region by using the template method and ab-initio prediction method respectively.By employing the parallel approach in the loop prediction, and caring about the inter-actions between loops during the process of the loops folding in parallel, our approachmakes the entire prediction process closer to the natural folding process. By usingthe pGPCR method to predict the3D-structure of two cases which are used in latestauthority testing, we achieve better results both in transmembrane and ECL2part ofthe result set. The testing results also indicate that our prediction results are closerto the natural structure in the loop region which contains the β fold and has stronginteractions between loops.In the study of GPCR’s side-chain prediction, this thesis improves the generatingmethod of GPCR’s side-chain searching space. And considering the infuences betweenside-chains, we propose a method of constructing rotamer library which based on afour level inference model. We conduct the side-chain prediction test experiments onthree types of GPCR protein, which confrms that the quality of the rotamer libraryis closer to that of the popular used rotamer library. And a new side-chain parallelprediction searching approach is applied. By parallelization of multiple ant colonies,each of which uses own energy function and parameters, we predicted the side-chain of pGPCR’s fve best results, and the results show that our method is competitive.