# GPU-Accelerated Virtual Screening and Molecular Dynamics Simulations for Identification of Novel DPP‑4 Inhibitors

## TL;DR

This paper uses GPU-powered simulations to find new, safer DPP-4 inhibitors for treating type 2 diabetes.

## Contribution

A novel GPU-accelerated computational pipeline for identifying DPP-4 inhibitors with high binding affinity and permeability.

## Key findings

- EPZ005687, OSU-03012, and bemcentinib showed higher binding affinity than FDA-approved drugs.
- These compounds exhibited better permeation characteristics across a membrane model.
- The computational strategy effectively identified promising antidiabetic candidates.

## Abstract

Inhibition of dipeptidyl
peptidase 4 (DPP-4) is a crucial therapeutic
strategy for the management of type 2 diabetes mellitus (T2DM). However,
current inhibitors often exhibit unwanted toxicity, underscoring the
need to discover novel, selective, and safer alternatives. This study
employs an integrated computational pipeline to accelerate the identification
of new DPP-4 inhibitor candidates. To that effect, GPU-accelerated
molecular docking of 30,699 bioactive PubChem compounds was combined
with molecular dynamics (MD) simulations and membrane permeability
analyses. A workflow that systematically filters candidates was presented
based on the score binding predicted by Uni-Dock. Subsequently, the
stability of 32 promising protein–ligand systems was assessed
using 100 ns MD trajectories, confirming their stable binding to the
DPP-4 active site. Compounds EPZ005687, OSU-03012, and bemcentinib
showed higher binding affinity and more favorable interactions within
pockets S1, S2, S1′, S2′, and S2 ′ than the FDA-approved
reference drugs like alogliptin, based on MM-GBSA calculations. To
assess the therapeutic viability of the candidates, their cellular
absorption potential was also investigated. Permeability (free energy
of transfer profile) and interactions were calculated via Umbrella
Sampling and long-time MD across a physiologically relevant enterocyte
membrane model. The results revealed that EPZ005687, OSU-03012, and
bemcentinib exhibited better permeation characteristics than alogliptin.
This combined evidence of high target affinity and enhanced cellular
permeability strongly suggests these compounds are up-and-coming antidiabetic
agents. These findings demonstrate the efficacy of this integrated
computational strategy, along with the utilization of rigorously filtered
public databases, for accelerating the discovery of safer and more
effective antidiabetic treatments.