Protein secondary sturcture prediction using artifical neural networks and support vectors machines

Abstract

In this thesis, a novel Hybrid Neural Networks Predictor (HNNP) system for the Protein Secondary Structure Prediction (PSSP) problem is described. By explor- ing such a new hybrid system, the intention is to investigate the feasibility of the HNNP in PSSP and even achieve improvements over existing methods. The proposed system is a cascaded network of the Radial Basis Function Neural Net- work (RBFNN) and the Multi-Layer Perceptron Neural Network (MLPNN).