Predictive model for protein function using modular neural approach

Abstract

As interest within bioinformatics has been vastly increased, efforts to predict functional role of proteins have been made using diverse approaches. In this paper, we discuss a protein function prediction method that utilizes protein molecular information including protein interaction data. The proposed method takes the given problem into account as a K-class classification problem and resolves the new problem by using a modular neural network based predictive approach. The simulation demonstrates that the proposed approach predicts the functional roles of Yeast proteins with unknown functional knowledge and is competitive to the other methodologies in KDD Cup 2001 competition. © Springer-Verlag Berlin Heidelberg 2005.