Extraction of prediction rules: Protein secondary structure prediction

Protein structure prediction is one of the most important problems in bioinformatics. Protein's secondary structure prediction is a key step in prediction of three-dimensional structure of protein. Despite all the efforts made, so far finding an accurate computational approach to solving a protein structure problem remains a challenging problem. Many computational techniques have been used to predict protein secondary structure (PSS) however, only few of such researches to have dealt with logic rules fundamental to prediction itself. This study, aims to combine decision trees at output of supporting vector machines (SVM) to extract rules governing protein secondary structure prediction. The rules share remarkable relations between the prediction model and the biological meaning. Moreover, they improved the intelligible of protein secondary structure prediction by providing an inside to the predicting model itself. Results revealed that the proposed rules were generated on RS126 data set, and these rules can also be explained biologically.