Signal Series-Wound Prediction of Protein Secondary Structure

Protein secondary structure prediction is a bridge between amino acid sequence and tertiary prediction. Various methods have been used to improve the prediction accuracy and have been developed greatly. Protein classification is a multi-class classification problem. For traditional method, the three structure are predicted in the same time. But it can be degraded to a set of binary classification problem, where one classifier is designed for each class. However, how to combine them again is a difficult question. We found that different prediction´s order can get different result. Here we propose a new method named signal series-wound method to predict the three structures signally by different order, and then by voting we get the best prediction result.This article propose a new method to predict protein secondary structure, to invalidate the effect of signal output. First, we predict the three structures separately, classifiers trained in different feature spaces.The proteins in one class are seen as negative example while those outside the class are seen as negative example. Then the three structure result are combined together by Series-wound method. The experiments on Database 396 show good result and prove the method is effectively.