I am searching for molecular docking software that exploit the acceleration of the graphical processor units (GPUs).
If anyone has some direct experience in using GPU based docking software, please share your experience.
I would like to have an idea about the CPU vs GPU performances (i.e. speed benchmarks for a virtual screening), but I am looking also for other things like the accuracy of the binding site prediction (in case of blind docking) or the ability to report the correct rankings in a virtual screening.
May I ask why you are looking for GPU to run docking simulations? If you are interested in any docking protocols working with GPU, you can have a look at the following papers.
Conference Paper GPU Accelerated Molecular Docking Simulation with Genetic Algorithms
Conference Paper Molecular Docking Simulation Based on CPU-GPU Heterogeneous Computing
https://www.semanticscholar.org/paper/GPU-Accelerated-Molecular-Docking-Simulation-with-Altuntas-Bozkus/41346b426eeb545783b1dd8d110d78612e01fae4
Look for litereatures to find out if there is any structural studies has been done. If yes, you can define the cavity or else predict the active site for the target using Pocketfinder, CASTp.
GOLD, GLIDE, MOE, Discovery Studio are commercial softwares for protein-ligand docking/ Virtual screening. Auto-dock is based on Lamarckian Genetic Algorithm, it is freely available. Don't use any online screening tools for docking/Virtual screening.
Try Experimental approaches like biochemical assays or else Molecular dynamics simulations to study the stability and interactions between the enzyme-inhibitor complexes.
Ramin Ekhteiari Salmas Thanks for your suggestions.
Since my group has a GPU cluster access but no CPU cluster I would prefer to exploit time consuming jobs (like Virtual screening of million of compounds) on the GPUs.
Running locally on my workstation (8xCPU) an Autodock Vina screening give me poor performances (about 5000 compounds per day) and it will take months to run a library of Million of compounds. So, I was wondering if some docking software have been optimized to speed up the Virtual screening on the GPUs and how about their accuracy of predictions.
Carmine, it’s very nice that you have gpu cores in your lab. Gpu is very helpful in running long molecular dynamics simulations and aren’t comparable with cpu in acceleration. However, gpu is not mostly used for docking simulations since the docking simulations aren’t based on common parallel protocols and only split the cpu cores to dock the ligands.
Hi Carmine Marco Morgillo
A git repository of Autodock is available as an OpenCL accelerated version that can run on GPU:
https://github.com/ccsb-scripps/AutoDock-GPU
Hope it helps
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence