Kernel based approach for protein fold prediction from sequence

Due to the relatively large gap of knowledge between gene identification and gene function, the ability to construct a computational model describing gene function from sequence information has become an important area of research. In order to understand the biological role of a specific gene, we will require knowledge of the corresponding protein's structure and function. We present a support vector machines based method for determining a protein's fold from sequence information alone where this sequence has little similarity with sequences with known structures. We have focused on improvement in multiclass classification, parameter tuning, descriptor design, and feature selections. The current implementation showed better performance than previous similar approaches.