Sequence Similarity and Support Vector Machine for Protein Secondary Structure

The majority of human coding regions have been sequenced and several genome sequencing projects have been completed. With the growth of large-scale sequencing data, an efficient approach to analyze protein is more important. Protein function and structure are foundations for drug design and protein-based product. However, it’s difficult to predict protein function and structure (three-dimension) directly from protein sequence. Therefore, analyzing protein secondary structure is indispensable. In the previous work, researchers always focused on classifying three states of protein secondary structure : helix, strand and coil classes. It’s a common classification problem for the prediction of protein secondary structure. Comparing with other machine learning methods for this problem, many studies usually ignore the protein local sequence/structure properties. It concerns the accuracy of prediction because there exists a large number of proteins that are homologous but whose sequences are only remotely related. In this thesis, we propose to use sequence similarity and Support Vector Machines (SVMs) to predict protein secondary structure. First, we try to encode the amino acids sequences( RS126 and CB513 ) and transform sequence segments into vectors for training. Second, we construct the SVM classifiers for classifying each residue of each sequence into the 3 secondary structure classes (i.e. H, E, or C). SVM has been successfully applied in pattern recognition problem. SVMs are learning systems that use a hypothesis space of linear functions in a high dimensional feature space, trained with a learning algorithm from optimization theory that implements a learning bias derived from statistical learning theory. It’s very suitable to compute with large-scale protein sequences. The challenges and comparisons of different approaches were demonstrated for protein secondary prediction.