gHPCSO: Gaussian Distribution Based Hybrid Particle Cat Swarm Optimization for Linear B-cell Epitope Prediction

Linear B-cell epitope (LBCE) identification is critical in developing peptide-based vaccines, antibody production, and immuno-diagnosis. Laboratory experiments are costly and time-consuming for this endeavour. Therefore, it is required to develop computational techniques to predict LBCE. Many techniques have been developed, but none of them achieved the highest accuracy due to high-dimensional LBC data. High dimensional data leads to computational complexity, and the inclusion of all the features may not provide an accurate prediction. An effective feature selection method is required to select the most prominent features from the high dimensional dataset. This paper presents a novel feature selection method for LBCE classification which is named as Gaussian distribution-based Hybrid Particle Cat Swarm Optimization (gHPCSO). The gHPCSO solves the problem of local optima and low convergence of Particle Swarm Optimization (PSO) by using seeking and tracing mode of Cat Swarm Optimization (CSO) where particle position is updated through CSO. The Gaussian distribution is employed for the population initialization of particles to improve the convergence. The benchmark dataset for the experiments is collected from the IEDB protein bank (LBtope Fixed, bCPred, ABCPred16, and Chen). Different feature extraction techniques are used to create feature vectors. These extracted features are provided to gHPCSO which uses k-nearest neighbour (k-NN) to classify LBCE and non-epitopes. Precision, recall, F-measure, Mathews correlation coefficient (MCC), and accuracy are considered for evaluation purposes. State-of-the-art approaches are compared with gHPCSO where results justifies the superiority of gHPCSO. Friedman's test is used to evaluate the consistency of the gHPCSO. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.