| Summary: | 碩士 === 國立臺灣大學 === 資訊工程學研究所 === 92 === This thesis reports a study on a two-stage protein structural alignment algorithm based on hyper-ellipsoidal clusters. The design of the two-stage algorithm is aimed at improving the efficiency of protein structural alignment without trading off analysis accuracy. In the first stage of the proposed approach, hyper-ellipsoidal clusters are employed to model the substructures of random coils as well as the ?helix and β-sheet structures. Due to this practice, the number of possible alignments of two protein tertiary structures to be examined in the first stage of analysis is substantially reduced and, as a result, the efficiency of the alignment operation is greatly improved. In the second stage of analysis, a refinement algorithm is invoked to fine-tune the alignment outputted by the first stage. The main distinction of the approach proposed in this thesis, in comparison with the existing approaches, is that the structural information of random coils is exploited so that the accuracy of analysis is not traded for efficiency. This thesis also reports the experiments conducted to evaluate the performance of the proposed approach. Experimental results reveal that protein structure alignment based on the hyper-ellipsoidal clusters generally achieves higher accuracy in both global alignment and local alignment than the existing algorithms.
|
|---|
