In Bid for Coronavirus Vaccine, U.S. Eases Access to Supercomputers
The U.S. government,
and others are giving researchers world-wide access to at least 16 supercomputers to help speed the discovery of vaccines and drugs to combat the novel coronavirus.
The White House Office of Science and Technology Policy on Sunday announced the Covid-19 High Performance Computing Consortium, a partnership that includes IBM, the Energy Department national laboratories,
Amazon Web Services,
Researchers from academia, government and the private sector will be able to submit coronavirus-related projects through a website beginning Sunday. Representatives from the consortium with backgrounds in areas such as high-performance computing, biology and epidemiology will approve the projects based on merit and the path to fastest impact, said Dario Gil, director of IBM Research.
The goal is to grant researchers access to powerful computing resources days after they submit their projects.
The announcement comes as health-care and government officials have criticized key parts of Washington’s response to the crisis, both in providing testing and helping secure medical supplies for hospitals.
“Ultimately, we need a cure. To be able to tackle that, we need to accelerate science,” Mr. Gil said. “We do have an extreme sense of urgency.”
The machines to which researchers will have free, remote access include two of the world’s most powerful supercomputers from IBM: Summit, at Oak Ridge National Laboratory, and Lassen, at Lawrence Livermore National Laboratory.
Combined, the 16 supercomputing systems that will be available to researchers at the start of the program represent over 265 petaflops of computing capacity, 775,000 computer processing unit cores, and over 34,000 graphics processing units, Mr. Gil said.
The machines were built by companies including IBM and others.
Supercomputers are used to tackle computational problems in initiatives such as drug discovery, materials development and battery design.
Such problems are complex and time-consuming to complete for standard computers. Predicting positive and negative effects of specific drugs, for example, is a computationally complex problem because it requires simulating the structure of molecules and their chemical features.
Supercomputers have thousands or tens of thousands of processors that coordinate and work together to perform large calculations, Mr. Gil said. Supercomputers are also particularly well-suited for processing and analyzing large amounts of data using artificial-intelligence algorithms, Mr. Gil said.
Researchers at the Energy Department’s Oak Ridge National Laboratory and the University of Tennessee recently used the IBM Summit supercomputer to run simulations on more than 8,000 compounds and identified 77 small molecules that could be used to accelerate the discovery of new therapies and vaccines for the coronavirus, said Paul Dabbar, undersecretary for science at the department.
Government officials began conversations with IBM about the initiative on March 18, said Michael Kratsios, the U.S. chief technology officer. Earlier last week, researchers released a vast data set for AI experts to mine. It contains more than 29,000 articles of scholarly literature about Covid-19 and related viruses. Since that data set was put out, it has been downloaded over 18,000 times by researchers who want to use AI algorithms to mine for insights, according to a White House official.
“Something we’ve seen as this pandemic has taken hold is the entire research community is ready, willing and eager to help assist the American people in this time of crisis,” Mr. Kratsios said.
Technology companies have also made other high-performance computing resources available for free to teams working to develop test kits and vaccines for Covid-19.
Microsoft and enterprise-technology firm Rescale Inc. announced such a program Thursday, while Amazon Web Services on Friday announced a $20 million initiative to accelerate Covid-19 diagnostic research.
Write to Sara Castellanos at email@example.com
Copyright ©2019 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8