Accelerating AutoDock4 with GPUs and Gradient-Based Local Search

Abstract

<div>AutoDock4 is a widely used program for docking small molecules to macromolecular targets. It describes ligand- receptor interactions using a physics-inspired scoring function that has been proven useful in a variety of drug discovery projects. However, compared to more modern and recent software, AutoDock4 has longer execution times, limiting its applicability to large scale dockings. To address this problem, we describe an OpenCL implementation of AutoDock4, called AutoDock-GPU, that leverages the highly parallel architecture of GPU hardware to improve the docking throughput up to 170-fold. Moreover, we introduce the gradient-based local search method ADADELTA, which is more efficient than the original Solis-Wets method, especially for conformationally complex ligands. We estimate a 1300x reduction in the number of scoring function calls for ligands with 20 rotatable bonds, and even higher reductions likely for more complex ligands. The improvements reported here, both in terms of docking throughput and search efficiency, expand the domain of applicability of the AutoDock4 scoring function considerably.</div>

Translational suppression of SARS-COV-2 ORF8 protein mRNA as a Viable therapeutic target against COVID-19: Computational studies on potential roles of isolated compounds from Clerodendrum volubile leaves

“…The novel compound (4,5-dihydroxy-6-(hydroxymethyl)-5-methoxy-3-(1,3,4,5-tetrahydroxypentan-2-yloxy)tetrahydro-2H-pyran-2-yloxy)-5-hydroxy-7-methyl-1,4a,5,6,7,7a-hexahydrocyclopenta[c]pyran-7-yl acetate) was constructed using the 2D-sketcher tool in Maestro software. To perform the molecular docking studies, the recent version of docking software, i.e., AutoDockGPU [ 48 ] together with its GUI AutoDockTools (ADT) [ 49 ], was used. Earlier, both the obtained 3D mRNA structures and all the ligands were prepared and converted into the AD4-PDBQT format utilizing the scripts prepare_ligand4.py and prepare_receptor4.py, which are a part of AutoDockTools.…”

Supercomputer-Based Ensemble Docking Drug Discovery Pipeline with Application to Covid-19

“…Cloud and distributed computing resources also provide this type of completely parallel solution for high-throughput screening 35,36 . The use of GPUs has recently been made possible for the widely-used program AutoDock4 37,38 resulting in the program AutoDock-GPU, which provides up to 50X speedup over AutoDock4 (available at https://github.com/ccsb-scripts/AutoDock-GPU) [39][40][41] . Thus, the use of leadership HPC facilities for ensemble docking can provide the ability to screen billions of ligands to a full set of conformations generated with HPC-based MD simulations.…”

Computational Study of C-X-C Chemokine Receptor (CXCR)3 Binding with Its Natural Agonists Chemokine (C-X-C Motif) Ligand (CXCL)9, 10 and 11 and with Synthetic Antagonists: Insights of Receptor Activation towards Drug Design for Vitiligo

“…A grid of 40 × 32 × 40 Å with spacing of 0.375 Å, centered on the residues identified as relevant for both receptor activation and agonist binding, was calculated with AutoGrid 4.2.6 [ 89 ]. Finally, molecular docking was performed with AutoDock 4.2.6 optimized for graphics-processing units, using a total of 25 runs and 25,000,000 evaluations, with a Lamarckian genetic algorithm and Solis–Wets local search [ 90 ] using three different conformations of CXCR3: the initial frame (F0) of the CXCR3-CXCL10 complex coarse-grained molecular dynamics simulation (CXCR3-CXCL10_CG-MD), the most representative conformation (C1) of the CXCR3-CXCL10_CG-MD simulation, and frame 7 (F7) corresponding to the 175-ns timestep of the 250-ns all-atom molecular dynamics (AA-MD) simulation of CXCR3-CXCL10.…”