Protein Secondary Structure Prediction Based on Denoeux Belief Neural Network

Predicting the secondary structure of protein is an important step towards obtaining its three dimensional structure and consequently its function. At present, the best predictors are based on machine learning techniques, in particular neural network architectures. We introduce a new architecture called Denoeux belief neural network (DBNN) for the prediction problem. DBNN uses reference patterns as items of evidence regarding the class membership of each input pattern under consideration. This evidence is represented by basic belief assignments (BBAs) and combined using the Dempster’s rule. DBNN has demonstrated excellent performance in other classification problems compared to existing statistical and neural network techniques. Our system, UTMPred incorporated the DBNN architecture and demonstrated a minimum protein secondary structure prediction accuracy, Q8 of 62.7% that is comparable to the best existing predictor, SSpro8 which has a classification accuracy of 62-63% range.