An evolutionary approach for protein classification using feature extraction by artificial neural network

A protein superfamily consists of proteins which share amino acid sequence homology and are therefore functionally and structurally related. Generally, two proteins are classified into the same class if they have most of the features extracted in common. As the size of the protein databases are becoming larger in size, it is better to develop an intelligent system to classify the protein with high accuracy. Artificial neural networks have been successfully applied to problems in pattern classification, function approximation, optimization, and associative memories. Multilayer feedforward networks are trained using the backpropagation (BP) learning algorithm but they are limited to searching for a suitable set of weights in an a priori fixed network topology. This mandates the selection of an appropriate optimized synaptic weight for the learning problem on hand. Genetic Algorithm (GA) is a stochastic based global searching technique which may be used to find out the optimized synaptic weight. Thus, a hybrid method combining GA-BP is implemented in this paper. Due to the limitations of GA such as premature convergence, low local convergence speed etc. an improvement to the GA is done. The Adaptive Genetic Algorithm (AGA-BP) uses an adaptive updating mechanism of the crossover and mutation probability which gives better result in comparison to GA-BP and traditional BP in terms of speed, predictive accuracy, and precision of convergence.