Hemoglobin secondary structure predicts with four kernels on support vector machines

Secondary structure prediction of proteins has increasingly been a central research area in bioinformatics. In this paper, support vector machines (SVM) are discussed as a method for the prediction of hemoglobin secondary structures. Different sliding window sizes and different kernels of SVM are comparatively investigated in terms of accuracy of prediction of hemoglobin secondary structure. For this purpose, the training and testing data were obtained from the Protein Data Bank, US with database of secondary structures of protein (DSSP). The results of prediction with different SVM kernels and different window sizes were found to be in the range of 5.93-15.90, 67.76-70.05 , 69.77-73.25, and 74.42-77.64 % for linear kernel, sigmoid kernel, polynomial kernel and Gaussian radial basis kernel, respectively