OpenMP API Helps Speed up Search for COVID-19 Drug

OpenMP API Helps to Speed Up the Search for a COVID-19 Drug

Author: Matthijs van Waveren

The Coronavirus Disease 2019 (COVID-19) has become a major threat worldwide due to its highly contagious nature. As of today, we have passed the mark of a million identified cases worldwide, and we have not yet reached the peak of the epidemic. At present, there are no established drugs available against it. The OpenMP API has been used to parallelize the Lamarckian Genetic Algorithm used in AutoDock 4.2, the de-facto standard application for molecular docking, one of the key steps in identifying drug candidates. This performance improvement should help us to find a drug sooner.

Molecular docking is one of the steps used in identifying drug candidates against SARS-CoV-2, the coronavirus that causes COVID-19. Molecular docking predicts whether a drug candidate can bind to SARS-CoV-2. It is one of the most frequently used methods in structure-based drug design, due to its ability to predict the binding conformation of small molecule ligands to the SARS-CoV-2 protein spikes. Characterization of the binding behavior plays an important role in the design of drugs.

AutoDock 4.2 simulates the molecular docking process between cells and the virus by predicting the ligand-receptor interactions. To perform a ligand-receptor docking experiment, the software accepts as inputs ligand and macromolecule coordinates, and then utilizes the Lamarckian Genetic Algorithm (LGA) to generate ligand positions and minimize binding energies. Each docking computation consists of multiple independent executions of the LGA. The LGA is a hybrid genetic algorithm with local optimization that uses a parameterized free-energy scoring function to estimate binding energy.

Using MPI and the OpenMP API, AutoDock 4.2 was parallelized for use on supercomputer systems [1] . AutoDock 4.2 was parallelized at multiple levels by:

1. Utilizing MPI to distribute AutoDock 4.2 docking experiments across a system,
2. Developing a grid map reuse scheme to reduce I/O, and
3. Implementing OpenMP parallelization of the Lamarckian Genetic Algorithm to achieve node-level parallelization.

The following table from [1] shows the LGA speedup achieved with the OpenMP API.

Threads Speedup % Ideal Serial 1.0 100 1 0.9 88 2 1.5 75 4 3.0 74 8 5.8 62 16 11.3 71 32 22.3 70

Using this parallelization, it is possible to increase the number of ligand-receptor docking experiments performed in a specific time, and thus to reduce the time to find a drug against SARS-CoV-2.

As an example, AutoDock 4.2 is being used on MOGON II, the supercomputer of the University of Mainz, in the prediction of candidate compounds as presumable SARS-CoV-2 inhibitors [2].

Speeding up drug discovery is urgently required and researchers around the world are working to identify novel drug candidates against COVID-19. AutoDock 4.2 is the de-facto standard for molecular docking, and it is at this very moment widely used by researchers [2, 3, and many others] in searching for COVID-19 drug candidates, as a literature search for COVID-19 and AutoDock will testify. This use case shows that the OpenMP API provides a useful and stable environment for node-level parallelization of AutoDock 4.2.

References

[1] Andrew Norgan, Paul Coffman, Jean-Pierre Kocher, David Katzmann and Carlos Sosa, Multilevel Parallelization of AutoDock 4.2, Journal of Cheminformatics 2011, 3:12
http://www.jcheminf.com/content/3/1/12

[2] Onat Kadioglu, Mohamed Saeed, Henry Johannes Greten & Thomas Efferth, Identification of novel compounds against three targets of SARS CoV-2 coronavirus by combined virtual screening and supervised machine learning. [Submitted]. Bull World Health Organ.
E-pub: 21 March 2020.
http://dx.doi.org/10.2471/BLT.20.255943

[3] Siti Khaerunnisa, Hendra Kurniawan, Rizki Awaluddin, Suhartati Suhartati, Soetjipto Soetjipto, Potential Inhibitor of COVID-19 Main Protease (Mpro) from Several Medicinal Plant Compounds by Molecular Docking Study,
https://www.preprints.org/manuscript/202003.0226/v1