Performance Analysis of Deep CNN Assisted Optimized HIV-I Protease Cleavage Site Prediction with Hybridized Technique

Abstract

In recent years, human immunodeficiency virus infection and acquired immune deficiency syndrome (HIV/AIDS) has emerged as a global health issue. The disease is caused by a virus that affects the CD4 cell in the human body that lowers the immune system in the human body. HIV-protease is the agent that replicates itself and affects the CD4 T cells in the human blood. To overcome the problem of replication, inhibitors can be analyzed and designed that can bind the active sites in the proteases. To design efficient protease inhibitors, the knowledge about the specificity of cleavage sites is essential. Several encoding techniques and classifiers have been proposed to study and analyze the active cleavage sites in proteases. This paper proposes a new model and comparatively analyses the performance of Hybridized SVM_Genetic modeling with Deep CNN assisted optimized prediction of Cleavage sites. For optimal tuning of activation functions, two metaheuristic algorithms such as moth search and dragonfly are proposed in this work. The performance of both the methodologies is compared based on different parameters such as accuracy, specificity, F1 score, sensitivity, and NPV. To authenticate the performance of the proposed model, standard data from machine learning algorithms called UCI repository is processed for experimentation. The performance measured is compared with existing available techniques for predicting cleavages.