Accelerating vaccine research for COVID-19 with high-performance computing and artificial intelligence
APRIL 28, 2020 • BLOG POST • PETER UNGARO, SENIOR VP AND GENERAL MANAGER, HPC & MISSION CRITICAL SOLUTIONS, HEWLETT PACKARD ENTERPRISE
IN THIS ARTICLE
- HPE is powering research to accelerate discovery of antibodies, antiviral agents and drug candidates to be tested for vaccine development
- U.S. DOE's Argonne National Laboratory, Lawrence Livermore National Laboratory and GENCI dedicate HPC and AI technologies to improve speed and accuracy of molecular dynamics simulations
Scientists worldwide are speeding up drug discovery to combat coronavirus with complex modeling, simulation, AI and machine learning capabilities
High-performance computing (HPC) is playing a leading role in our fight against COVID-19 to support the urgent need to find a vaccine that will save lives and reduce suffering worldwide.
Scientists in labs around the globe rely on the massive computing power of HPC and supercomputers to run complex mathematical models, which transform vast volumes of evolving COVID-19 data into simulations of biological and chemical processes. These simulations advance our understanding of the new strain of virus and the complex interactions of the human body down to the molecular level, to accelerate the development of new treatments and preventative measures.
By combining modeling and simulation capabilities with new techniques in artificial intelligence (AI) and machine learning, these simulations are now becoming even more accurate. That is why through our collaborations with worldwide leading research centers that are using our HPC and AI solutions, Hewlett Packard Enterprise and the HPC industry are supporting scientists tackling complex research that will unlock insights and bring us closer to drug discovery.
These initiatives, which represent both public and private sector investments, are driven by a global network of scientists, researchers, AI and supercomputing specialists, who in many cases are working together to share data, collaborate, and make vital contributions in the overall effort.
HPE is proud to play a leading role in supporting many of these initiatives with HPE HPC systems and expertise. For example, the U.S. Department of Energy’s Argonne National Laboratory (ANL) and Lawrence Livermore National Laboratory (LLNL), and France’s National Center for Scientific Research (CNRS) together with GENCI, the French national infrastructure for HPC resources and facilities, are using HPC and AI to speed up discovery of antibody and drug candidates that can be tested for new vaccine treatments.
Each of these research teams are separately applying AI and machine learning to modeling and simulation to increase accuracy and predictions – something we describe as the emerging convergence of modeling and simulation with AI and analytics. These efforts further accelerate discovery of new antibodies, which are blood proteins produced to fight toxins or other foreign substances that induce immune response, or potential drug candidates based on existing catalogs of data. Researchers are then able to test these for potential counter-measures to the virus that can be developed into drug therapies.
Argonne researchers apply AI-enabled modeling and simulation to significantly speed discovery of antiviral agents
At the U.S. Department of Energy’s (DOE) Argonne National Laboratory, researchers have taken on a mission to dramatically accelerate the pace in discovering antiviral agents to counterattack the new virus, reducing the potential timeframe from years to just couple of months.
Researchers are using the Theta supercomputer, powered by HPE, and housed at the Argonne Leadership Computing Facility, to apply artificial intelligence and machine learning to accelerate the process of simulating billions of different small molecules from a publicly available database of drug candidates. The goal is to improve predictions on how molecules in drug candidates interact with each other and bind to viral proteins. Successful binding means these drug candidates can be used for further testing for a vaccine treatment.
Arvind Ramanathan, a computational biologist in Argonne’s Data Science and Learning division with whom we work, characterized the opportunity this way:
“When we’re looking at this virus, we should be aware that it’s not likely just a single protein we’re dealing with — we need to look at all the viral proteins as a whole. By using machine learning and artificial intelligence methods to screen for drugs across multiple target proteins in the virus, we may have a better pathway to an antiviral drug.”
LLNL uses a first-of-its-kind AI-driven modeling platform to design 20 initial antibody candidates among 1040 possibilities
We are incredibly proud to share that LLNL has already made significant progress narrowing down the number of potential antibody candidates from 1040 to an initial set of just 20! That’s a dramatic process of elimination. On top of that, this inspiring breakthrough was achieved in just weeks, compared to a typical lead time of years using other approaches.
LLNL’s COVID-19 response team, which includes researchers from various disciplines with deep expertise in vaccine and countermeasure development, used LLNL’s Catalyst, an HPC cluster powered by HPE, to improve predictions and speed up this discovery process by using a first-of-its-kind modeling platform. The platform integrates important components to generate high-quality predictions, such as experimental data and structural biology data, with bioinformatics modeling, molecular simulations and machine learning algorithms.
“Our approach, while still being developed, is aimed at designing high quality antibody therapeutics or vaccines in extremely rapid time-scales for scenarios in which waiting for many rounds of time-consuming experimental steps is not an option. Experimental data and structural bioinformatics are important components to enable high-quality predictions, but integrating machine learning and molecular simulations on HPC are the key to enabling the speed and scalability we need to search and evaluate huge numbers of possible antibody designs.” – Dr. Daniel Faissol, data scientist at Lawrence Livermore National Laboratory
CNRS and GENCI offer a faster approach to perform molecular dynamics simulations
Similarly, GENCI, together with CNRS, has extended its Jean Zay supercomputer, which is installed in IDRIS, one of CNRS’ data centers, to power converged modeling, simulation, AI and machine learning workloads, to a number of research centers also focused on increasing accuracy and outcomes of antibodies. GENCI’s Jean Zay was designed by HPE in direct response to President Emmanuel Macron’s significant AI initiative to propel France’s R&D in AI with a new supercomputer.
As with many other brilliant and diligent researchers responding to COVID-19, the team using GENCI’s Jean Zay, led by Pr. Jean Philip Piquemal at Sorbonne University in Paris, is taking a unique and innovative approach to deciphering and understanding the new strain of virus internal machinery to help increase discovery of new drugs. While still in an early research stage, Dr. Piquemal and his team are using the system to optimize the Tinker-HP software, a new approach to using parallel computing that is enabled by multiple GPUs and specifically designed to simulate at the level of atoms for large biological molecules. Using advanced molecular dynamics techniques captures many-body physics, and Tinker-HP helps scientists achieve this by simultaneously performing a number of data-intensive calculations to create 3D simulations of molecular interactions faster and at a higher resolution.
HPE is proud to be a force for good in even the most trying of times
Being a part of a global community that is making a deep, meaningful impact and helping to save lives – particularly in this time of global urgency – gives me tremendous pride.
At HPE, we are committed to advancing the way we live and work. As a world leader in HPC and AI, we recognize the impact we can make by applying modeling, simulation, machine learning and analytics capabilities to data to accelerate insights and discoveries that were never before possible. By making these technologies available and setting new standards for speed, performance and scale, we hope to enable the broader HPC research community to make scientific breakthroughs related to the COVID-19 pandemic that will advance treatment and save lives.
We are extending our HPC efforts to address COVID-19 research on a greater, national scale by recently partnering with the White House as part of the COVID-19 High Performance Computing consortium. We are further helping to address other challenges with a number of initiatives that can be found in the HPE COVID-19 microsite and latest blog from Antonio Neri, HPE President and CEO.
Thank you all, and be sure to continue practicing social distancing and washing your hands!