An Efficient Classification of Tuberous Sclerosis Disease Using Nature Inspired PSO and ACO Based Optimized Neural Network

Abstract

Tuberous sclerosis disease is a multi-system genetic disorder that broadly affects the central nervous system resulting in symptoms including seizures, behavior problems, skin abnormalities, kidney disease etc. This hazardous disease is caused by defects, or mutations of two genes-TSC1 and TSC2. Hence, analysis of TSC1 and TSC2 gene sequences can reveal information which can help fighting against this disease. TSC2 has 45 kilobases of genomic DNA, 41 known exons, and codes for a 5474-base pair transcript. On the other hand, the TSC1 gene spans about 53 kb of genomic DNA with 23 exons coding for hamartin, a hydrophilic protein with 1164 amino acids and 130 kb DNA. It is not possible to manually extract and analyze all the hidden information lies in TSC1 and TSC2 in wet lab. Machine learning approaches have been extensively applied to discover hidden information lies in any dataset. Efficient machine learning approaches need to be discovered to analyze TSC1 and TSC2. This chapter concentrates on using convolutional neural network optimized with nature inspired approaches such as, Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO). The main challenge of any machine learning approaches is to optimize its parameters effectively. Since, both PSO and ACO are iterative optimization approaches to iteratively develop a candidate solution it can be employed with machine learning approaches to reduce the effort to formulize the parameters of various machine learning techniques. Besides, all the weights, biases, learning rates are optimized with PSO and ACO algorithms. The proposed approach has been implemented for classification of tuberous sclerosis disease. Additionally, for comparison purpose we have employed decision tree, naïve bayes, polynomial regression, logistic regression, support vector machine, random forest etc. Comparative analysis of time and memory requirements of all the approaches have been performed and it is found that the efficiency and time requirements of proposed approach outperform its competitors. Meanwhile, Apriori algorithm have been applied to generate association rules and to extract effective information regarding dependencies of attributes with each other to identify those that are responsible to empower this disease. For exploration and extraction of different significant information from the sequence of TSC1 and TSC2 gene, frequent itemset along with analysis of mutation sequence combining other approaches have been illustrated. Statistical analysis on the same dataset reveals the similar findings as the Apriori algorithm.