Computational Physics (MELD) modeling for drug discovery: SARS-CoV-2
To accelerate drug discovery for the SARS-CoV-2 virus, using computational Physics (MELD) modeling. An accelerator for Molecular Dynamics (MD) modeling. (MELD x MD accelerates MD modeling, often by orders of magnitude. It folds proteins, binds ligands and refines structures and protein-protein docking.
Finding safe and effective anti-viral agents is hard. But it's important and timely for the Covid-19 disease.
First, viruses have far fewer proteins (a few tens, vs. a few thousand in bacteria) that a drug research campaign could target. Second, viruses use the human cell's own machinery to reproduce, so you have to be careful that the drug you design does not kill the host. Third, viruses mutate rapidly as they pass through human populations, so they can quickly "learn" how to become resistant to your drug. And, finally, drug discovery is always hard, no matter what the disease.
Computer modeling in Drug Discovery:
Computer modeling is often a front-line design tool. Computational modeling helps discover drugs faster and of higher quality. To use it, you need to know the protein structure, i.e. where all the atoms are located on the viral protein you want to attack. With modern technologies like cryo-EM, protein structures can be determined very quickly. Virtual screening methods (algorithms like DOCK) are already being applied; it’s too early to know the results. These methods search through large databases of compounds, to find ones worth testing in the lab. The shortcoming is that virtual screening is just an empirical computational filter. It only tells you the hundreds to thousands of molecules that might work.
Physics based Computer modeling:
Rather than empirical methods, the state-of-the-art is now computationally-intense physics-based molecular simulations. This modeling is not just large-scale filtering. It has two advantages: First, it can help medicinal chemists understand the biological mechanism of how the virus is causing damage in the first place. Second, in principle it can pinpoint more precisely which drug candidates might be the very best ones. It does this by computing the drug binding affinity to the protein target. To run these simulations, we used High Performance Computing (HPC) Notebook with underlying infrastructure powered by 128 Nvidia RTX 6000 GPUs, orchestrated by Zeblok’s AI PaaS with MPI. The state-of-the-art orchestration helped distribute the AI-workload on hundreds of GPUs with few clicks.
Zeblok AI- Platform resources used:
Zeblok Computational platform
HPC Notebook with MPI enabled
Multiple containers to support multi-GPU, multi-CPU compute engines
128 RTX6000s GPUs
50GB Block Store
500GB of Parallel File System
About The Laufer Center
The Louis and Beatrice Laufer Center, established in 2008 to advance biology and medicine through discoveries in physics, mathematics and computational science, is a hub for Physical and Quantitative Biology research at Stony Brook University. Laufer Center researchers come from several Stony Brook departments and Cold Spring Harbor Laboratory. For more information: