Protein secondary structure prediction based on improved SVM method in compound pyramid model

Methods for predicting protein secondary structure provide information that is useful both in ab initio structure prediction and as additional restraints for fold recognition algorithms. Secondary structure predictions may also be used to guide the design of site directed mutagenesis studies, and to locate potential functionally important residues. In this article, we propose a method of improved SVM for predicting protein secondary structure. Using evolutionary information contained in amino acid´s physicochemical properties, position-specific scoring matrix generated by psi-blast as input to improved SVM, secondary structure can be predicted at significantly increased accuracy. Based on KDTICM theory, we have constructed a compound pyramid model, which is composed of four layers of the intelligent interface and integrated in several ways, such as improved SVM, mixed-modal BP, KDD* method and so on. On the RS126 data set, state overall per-residue accuracy, Q3 reached 83.06%, while SOV99 accuracy increased to 80.6%.On the CB513 data set, Q3 reached 80.49%, SOV99 accuracy increased to 79.84%.This article briefly introduces this model and highlights the improved SVM method.