A specialized Anton 2 supercomputer, developed by D. E. Shaw Research and hosted at PSC, is now available for urgent research with the potential to impact the national COVID-19 response, thanks to agreements between D. E. Shaw Research, PSC and its funding organizations, and collaborating centers. The availability of this unparalleled resource for simulating the motions and interactions of large biomolecules with each other and with smaller molecules such as candidate drugs could contribute to understanding of the virus that causes COVID-19 and potentially help accelerate development of treatments, preventives, and possibly vaccines.
Anton was developed by D. E. Shaw Research to execute molecular dynamics simulations of biomolecules (e.g., proteins, nucleic acids, and lipids) orders of magnitude faster than was previously possible. Anton’s speed allows scientists to simulate these complex biomolecular systems, and their interactions with natural signaling molecules and drug candidates, over ultra-long timescales, enabling observation of important biological events. The Anton 2 system hosted at PSC has been provided without cost by D. E. Shaw Research for non-commercial use by the U.S. research community, and PSC supports the community’s use of this resource with operational funding from the National Institutes of Health.
Applications for computer time on the Anton 2 at PSC are being accepted through the COVID-19 HPC Consortium. The consortium also allocates COVID-19 research time on PSC’s large Bridges supercomputing platform as well as on other consortium members’ computers.
A Carnegie Mellon University team is using artificial intelligence (AI) on PSC’s Bridges-AI platform to investigate possible candidates to fight the virus that causes COVID-19. As part of PSC’s membership in the COVID-19 High Performance Computing Consortium, a group led by Prof. Olexandr Isayev of the CMU Department of Chemistry is combining AI with computational chemistry using the AI-specialized GPU nodes on Bridges-AI. Their unique approach may be as much as a million times faster than the usual quantum mechanical calculations needed for such simulations.
The CMU scientists aim to apply AI-accelerated simulations to aid the design and repurposing of antiviral drugs specific to the virus. To achieve this goal, they must first understand the interactions of drugs and other small molecules with the virus’s proteins. Their step-wise approach will first analyze via AI a library of molecules that can be purchased from chemical companies, preparing them for automated screening in the computer. The best candidates from this screening will then be simulated on the supercomputer using a method called molecular dynamics, also enhanced with AI. Top hits from that final screening will be tested in partners’ laboratories.
The project presents an efficient strategy to discover new antiviral drugs with higher accuracy and potentially reduced cost compared with traditional drug discovery. Isayev’s work also raises the possibility of broader applications to real-world drug discovery pipelines. All screening datasets will be deposited to the COVID-19 Data Lake database, so other researchers can benefit from them as soon as possible.
Isayev has also incorporated COVID-19 research into his Special Topics in Computational Chemistry: Machine Learning course at CMU’s Mellon College of Science.
Bridges is funded by the National Science Foundation. PSC offers the platform to the open research community at no cost to scientists. Information on obtaining computing time on PSC and other systems via the COVID-19 High Performance Computing Consortium can be found here.
Investigators at the Pennsylvania State University have begun using PSC’s Bridges platform to study whether people’s use of Twitter can provide clues as to the course and severity of the COVID-19 pandemic. The goal is to extract information—such as whether tweets about COVID-associated symptoms could be a warning sign of an outbreak—that will help public health and government decision makers determine when and what responses can contain spread of the virus.
The project, led by Guangqing Chi of the Penn State Computational and Spatial Analysis Core, involves collecting and analyzing a massive amount of publicly available Twitter data in a given region using Bridges and Penn State’s ICDS Advanced CyberInfrastructure supercomputer. An aspect of Bridges that proved important to the work was its Hadoop-friendly nodes. Hadoop, a popular cloud-based tool for analyzing data, will not run on most supercomputers. Bridges allowed the researchers to scale up Hadoop tools they had developed on smaller computers by seamlessly moving them to the PSC resource.