Supercomputing Pipelines Search for Therapeutics Against COVID-19

Abstract

The urgent search for drugs to combat SARS-CoV-2 has included the use of supercomputers. The use of general-purpose graphical processing units (GPUs), massive parallelism, and new software for high-performance computing (HPC) has allowed researchers to search the vast chemical space of potential drugs faster than ever before. We developed a new drug discovery pipeline using the Summit supercomputer at Oak Ridge National Laboratory to help pioneer this effort, with new platforms that incorporate GPU-accelerated simulation and allow for the virtual screening of billions of potential drug compounds in days compared to weeks or months for their ability to inhibit SARS-COV-2 proteins. This effort will accelerate the process of developing drugs to combat the current COVID-19 pandemic and other diseases.

FIGURE 1.

SARS-CoV-2 virion visualized by Thomas Splettstoesser scistyle.com under commission. Used with permission.

FIGURE 2.

Docking example, emphasizing how binding pockets in a protein (left, with secondary structure shown as a green cartoon below a semitransparent surface), in this case, the SARS-CoV-2 Main Protease, can be occupied by small molecules. Two alternative bound molecule configurations are shown, black representing a crystallographic structure, and yellow representing the pose predicted by AutoDock-GPU.

FIGURE 3.

Graphical representation of the structural ensemble generated from parallel-tempering MD simulations. Each of the 26 protein snapshots taken from the ensemble has a different color, and is drawn in a cartoon representation. The spaghettilike strands represent loop elements where the greatest variation in structure takes place in the simulation.

FIGURE 4.

Overview for the workflow, starting from the small molecule database of potential drugs (1) and ending at a query-able database of results (9). Preprocessing the small molecule database (2) creates files needed for docking, and in our case was done on a small commodity cluster (CADES). Once on the file system, the docking workflow is prepared (3), and is stored within the slate–marble database system. Subsequently, the input files are loaded onto the fast NVMe drives on Summit (4), and the batches of compounds stored within the workflow database (5) are docked on Summit (6). Output is then converted to Apache parquet format (7) for storage into a GPU-accelerated Blazing SQL database (8). Queries on this database are fast and enable potential machine learning applications (9).

FIGURE 5.

Example of the Jupyter notebook interface for exploring the billion compound docking results.