Machine learning-assisted and simulation-based elucidation of phytochemical dipeptidyl peptidase-4 inhibitors towards...

A global problem with Diabetes Mellitus (type 2) is the disruption of insulin release as well as glucose regulation. The Dipeptidyl Peptidase-4 (DPP-4) is responsible for breaking down incretin hormone; therefore, it has an indirect effect on decreasing insulin release. Synthetic inhibitors of DPP-4 are clinically effective; however, due to the many negative side effects associated with them, alternative options are being explored that are much safer. Using an integrated computational pipeline combining cheminformatics-driven chemical space analysis, machine learning-assisted rescoring, ADMET filtering, molecular docking, molecular dynamics (MD) simulations, and systems-level pathway analysis to identify promising phytochemicals as DPP-4 inhibitors. A library of 250 phytochemicals that had been identified as possessing antidiabetic properties was obtained from public databases and then examined using UMAP and PCA so that structural diversity could be determined. Distinct clusters of similar chemical structures were identified using this analysis, and the top 20 compounds identified underwent evaluation of their ADMET properties through the use of SwissADME, pkCSM and admetSAR to confirm that they exhibited appropriate pharmacokinetic and toxicology profiles. Through protein docking with DPP-4 (PDB ID: 6B1E), coumarin, eugenol and cinnamic acid were highlighted as having a high binding affinity when compared to the known inhibitor Vildagliptin. The stability of these complexes was supported by using 500 ns MD simulations and MM/PBSA calculations. In addition to the molecular information, the results of the protein interaction and pathway enrichment analysis indicated that DPP-4 has many functions related to the regulation of insulin secretion and inflammatory processes through a complex network. Therefore, these studies support the use of phytochemical compounds as safe DPP-4 Inhibitors, and offer a new approach for rapid identification of plant-derived compounds for the treatment of T2DM.