Multiobjective Approach to Prediction of Protein Subcellular Locations

This chapter presents a comprehensive method for predicting protein subcellular location by employing a multiobjective particle swarm optimization-based feature selection technique. Utilizing the general form of pseudo-amino acid composition, the construction of protein features from sequences is detailed, with data representation involving a protein sample versus amino acid compositions as features. The algorithm’s objective is to simultaneously maximize feature relevance and minimize redundancy. Upon application to a multiclass dataset, the algorithm identifies relevant features. Comparative performance analysis against single-objective counterparts, such as sequential forward search, sequential backward search, minimum redundancy maximum relevance, Fisher discriminant, and a Cluster-based technique, underscores the effectiveness of the proposed method. This holistic approach addresses the intricacies of predicting subcellular protein localization, highlighting the multiobjective optimization’s prowess in refining feature selection for enhanced prediction accuracy.