Protein secondary structure prediction with high accuracy using Support Vector Machine

Abstract

Mining bioinformatics data is an emerging area of research. Proteomics is one of the largest areas of focus in bioinformatics and data mining research. Protein structure prediction is one of the most crucial and decisive problem in all the areas of research. Protein secondary structure can be used for the determination of the tertiary structure via the fold recognition method. Hence, predicting the secondary structures from the proteinpsilas primary sequences has attracted the attention of many researchers. Experimental methods have proved to be complex and expensive. So to develop a simple and accurate method for structure prediction is of great importance. In this paper, a new method has been proposed based on the machine learning technique. The first step of this proposal is to find out frequent patterns of consecutive amino acids in a protein database. After this, a set of frequent words (feature set) is found. Then support vector machine (SVM) is used as a binary/tertiary classifier for the classification of protein secondary structure with these frequent words.