COVID-19 main protease inhibitor design

Thomas Cheatham, University of Utah

Usage Details

In 2013, our research group was funded to develop a simulation and analysis workflow to design novel D-peptides that bind to a protein complex from Ebola that blocks membrane fusion  The workflow, based on known crystal structures, uses Rosetta to select different amino acid side chains on the fixed peptide backbone template and then performs 0.5- 1.0 microsecond-scale unrestrained molecular dynamics simulations with AMBER on GPUs (with modern force fields in explicit solvent) to optimize the structures, which are then ranked based on a simple MM-PBSA free energetics estimate.

The best candidates are then put through the workflow again, and this leads to optimization of the pool. Subsequent rounds of refinement shift the distribution to even lower free energy structures representing better binding. 

The workflow has been developed and automated, with analysis results populating a SQLite database and custom Python scripts were developed to pull data from Rosetta, AMBER, CPPTRAJ, and the MM-PBSA calculations. The AMBER GPU engine is very efficient on GPU resources running effectively 10 times the performance of the CPU code, with simulations of 300 ns per day possible on a single V100 GPU for systems of this size. 

The crystal structure of the COVID-19 main protease in complex with an inhibitor N3 was recently “published”  and dubbed as structure 6LU7. The inhibitor is a peptide. Our proposal is to apply the Ebola peptide design workflow to this system with the intention of getting preliminary data for planned future COVID-19 grant submissions. 

So, the idea is to use the short allocation window to further validate the workflow and to apply it to this COVID-19 main protease. As the workflow is already developed our team is ready to utilize resources immediately and our group has considerable HPC experience on NSF resources, including XSEDE and Blue Waters