Deep_Hybrid_CPP: Robust and Novel Deep Hybrid Learning Approach for Identification of Cell Penetrating Peptide using Multiview Feature Fusion

Abstract

CPPs are short peptide sequences that have the ability to pass through cell membranes and deliver a range of molecular cargoes within cells. Their remarkable ability to enter cells without inflicting significant harm to the membrane makes them crucial for uses like as intracellular imaging, targeted drug delivery, and gene therapy. I introduced Deep_Hybrid_CPP, an innovative computational approach for efficiently and effectively identifying cellpenetrating peptides through a multi-view feature fusion framework. This study integrates 10 diverse sequence-based feature extractors from different perspectives with 12 prominent machine learning (ML) algorithms in Deep_Hybrid_CPP to create multi-view features that thoroughly represent the essential information of cell-penetrating peptides. To enhance the distinguishing capability of my tailored genetic algorithm, Additionally employed it to select a collection of multi-view features. Based on a series of comparative experiments, my multi-view features outperformed certain traditional feature extractors regarding their discriminative capabilities. Additionally, regarding the independent test dataset, Deep_Hybrid_CPP achieved the top accuracy (ACC) and Matthew's correlation coefficient (MCC) of 97% and 94%, respectively, showing increases of 5.06% and 10.97%. Using this study's innovative computational approach, I expect to effectively evaluate and prioritize candidate peptides that may demonstrate favorable cell-penetrating peptide characteristics