Machine learning assisted screening framework for insecticidal peptides

Pest management in agriculture relies on the use of chemical pesticides. These chemicals are harmful to both the environment and to human health. This has necessitated the search for sustainable alternatives. In this regard, natural products and "biopesticides" are promising for sustainable crop protection. A handful of proteins and peptides with insecticidal properties have been extracted from different organisms till date. However, experimental screening and identification of insecticidal peptides is laborious and resource-intensive. In this study, we have developed a computational screening framework that can aid in the identification of insecticidal peptides. We have used non-linear support vector machine (SVM) algorithm to develop a binary classifier that can detect whether a peptide sequence is likely to have insecticidal property. To develop the binary classifier model, we have implemented three feature representation techniques namely, amino acid compositions (AAC, 20-dimensions), composition transition and distribution (CTD, 147-dimensions), and conjoint triad feature (CTF, 343-dimensions). Results obtain indicate that the CTF method outperforms the other two methods with an accuracy of about 0.947. Thus, the proposed approach can be used as a tool for sustainable agriculture.