Deep learning-based dipeptidyl peptidase IV inhibitor screening, experimental validation, and GaMD/LiGaMD analysis

Abstract

Abstract Background Dipeptidyl peptidase-4 (DPP4) is considered a crucial enzyme in type 2 diabetes (T2D) treatment, targeted by inhibitors due to its role in cleaving glucagon-like peptide-1 (GLP-1). In this study, a novel DPP4 inhibitor screening strategy was developed, which significantly improved screening accuracy. Results In this study, a DPP4 inhibitor screening method was developed, integrating receptor-based ConPLex, ligand-based KPGT, and molecular docking to enhance screening accuracy. Using this approach, four potential drugs were identified from the FDA database, achieving a 100% hit rate. Among these, Isavuconazonium demonstrated the highest inhibitory activity (IC50 = 6.60 µM). Furthermore, a user-friendly server, DPP4META, was established to predict IC50 values for DPP4 inhibitors. The binding and dissociation mechanisms of these drugs with DPP4 were further examined through Gaussian accelerated Molecular Dynamics (GaMD) and ligand Gaussian accelerated Molecular Dynamics (LiGaMD), revealing strong correlations with IC50 values. Additionally, a Python-based toolkit, pymd, was developed to facilitate protein-compound binding analysis. Conclusions Our study offers a robust approach and valuable insights for the development of DPP4 inhibitors, providing an effective means to investigate the binding and dissociation mechanisms between proteins and compounds.

Keywords

• DPP4 inhibitor
• Molecular dynamics simulation
• Protein-ligand interactions