PhytoAFP: In Silico Approaches for Designing Plant-Derived Antifungal Peptides

Abstract

Emerging infectious diseases (EID) are serious problems caused by fungi in humans and plant species. They are a severe threat to food security worldwide. In our current work, we have developed a support vector machine (SVM)-based model that attempts to design and predict therapeutic plant-derived antifungal peptides (PhytoAFP). The residue composition analysis shows the preference of C, G, K, R, and S amino acids. Position preference analysis shows that residues G, K, R, and A dominate the N-terminal. Similarly, residues N, S, C, and G prefer the C-terminal. Motif analysis reveals the presence of motifs like NYVF, NYVFP, YVFP, NYVFPA, and VFPA. We have developed two models using various input functions such as mono-, di-, and tripeptide composition, as well as binary, hybrid, and physiochemical properties, based on methods that are applied to the main data set. The TPC-based monopeptide composition model achieved more accuracy, 94.4%, with a Matthews correlation coefficient (MCC) of 0.89. Correspondingly, the second-best model based on dipeptides achieved an accuracy of 94.28% under the MCC 0.89 of the training dataset.Emerging infectious diseases (EID) are serious problems caused by fungi in humans and plant species. They are a severe threat to food security worldwide. In our current work, we have developed a support vector machine (SVM)-based model that attempts to design and predict therapeutic plant-derived antifungal peptides (PhytoAFP). The residue composition analysis shows the preference of C, G, K, R, and S amino acids. Position preference analysis shows that residues G, K, R, and A dominate the N-terminal. Similarly, residues N, S, C, and G prefer the C-terminal. Motif analysis reveals the presence of motifs like NYVF, NYVFP, YVFP, NYVFPA, and VFPA. We have developed two models using various input functions such as mono-, di-, and tripeptide composition, as well as binary, hybrid, and physiochemical properties, based on methods that are applied to the main data set. The TPC-based monopeptide composition model achieved more accuracy, 94.4%, with a Matthews correlation coefficient (MCC) of 0.89. Correspondingly, the second-best model based on dipeptides achieved an accuracy of 94.28% under the MCC 0.89 of the training dataset.