Analysis and comparison of the kernel accuracy for saccharomyces genus protein sequence classification

Even as the modern protein sequencing technology is uncovering great amount of new protein, such data are futile without knowledge of the functions the protein sequences encode. To overcome the limitation of experimental analysis, utilizing Support Vector Machine and kernel methods for functional prediction of unannotated protein has become a promising topic of research in the field of computational biology, inducing many researchers to develop kernels with improved accuracy and efficiency. In this paper, we assigned the Gaussian, Polynomial, and Normal kernels to each Sensu Stricto, Sensu Lato, and Petite-Negative groups of the Saccharomyces fungus species, for which the kernel showed the greatest accuracy in protein sequence classification. From the result we discovered the sequential shapes of the proteins and detected similarities of the structural linearity among the proteins species belonging in the same group. The resulting data allow us to provide an important categorization of kernels that will predict protein function with the greatest accuracy depending on the group of the Saccharomyces the protein belongs for future researches using sequential analysis for prediction, into which you can type your own text.