An approach of protein secondary structure prediction based on homology analysis method in compound pyramid model

In this paper, we propose a method of homology analysis for predicting protein secondary structure. The BP network model is based on pair-wise sequence alignment, and the accuracy of the prediction is improved. In the modeling stage, we use penalty function in the pair-wise sequence alignment and the result of alignment is put into the training set of the BP network. Considering the information of neighboring residues of the protein sequence, we use the BP network in modeling stage. At last, we apply this homology analysis method to the homology analysis module located at integrative layer of compound pyramid model proposed by us and get better results in our experiment comparing with other methods.