Welcome to the Spring 2019 issue of PSC Science Highlights!
Over the last six months we have been busy upgrading PSC’s capabilities in Big Data and artificial intelligence (AI). Arguably our most significant event was delivery of the first instance of the NVIDIA DGX-2 enterprise research AI system available to open research.
The DGX-2 is part of Bridges-AI, a $1.8-million, NSF-funded upgrade to our Bridges system that has added 88 NVIDIA Volta GPUs to an already successful CPU/GPU AI platform. Among other users, our Open Compass program will leverage Bridges-AI to help AI researchers across the nation tackle the most difficult problems in medicine, security, energy usage, business and other fields.
With the CMU Libraries, we are also organizing the first Artificial Intelligence for Data Discovery and Reuse conference at Carnegie Mellon. AIDR2019, to be held in May, will provide a platform for AI/machine learning researchers, data professionals and scientists to address data challenges and facilitate the next breakthroughs in science and technology using the power of AI and scientific data.
Featured in this issue is a Pennsylvania State University group that used Bridges’ CPUs for machine-learning recognition of impending extreme weather. Next, they hope to leverage the deep-learning power of Bridges-AI to improve on their results. In a very recent AI success, the Libratus imperfect-information-game-playing AI designed at the CMU School of Computer Science (and which, as you have read in previous issues, runs on Bridges), has brought $10 million in Department of Defense funding to Pittsburgh.
Along the other axis of PSC’s major emphasis—data science—is a new collaboration with the Compact Muon Solenoid (CMS) experiment at Fermilab. You can read about their use of Bridges to simulate possible signals in the “High-Luminosity LHC,” a 10-fold power upgrade to the Large Hadron Collider.
In bioinformatics, PSC received two major grants. PSC and the Department of Biomedical Informatics at the University of Pittsburgh’s School of Medicine have received an initial $598K in NIH funding as part of the Human BioMolecular Atlas Program (HuBMAP). Another three-year $289K NSF grant will further our work as part of the National Center for Genome Analysis Support (NCGAS). This issue features a genomics story in which a UC Davis team used Bridges to identify genes associated with surviving climate change in birds.
Two molecular dynamics stories in this issue leveraged two different PSC systems with complementary strengths. A University of Delaware group used Bridges’ ability to simulate extremely large molecular assemblies to uncover how a monkey protein prevents infection by HIV. A Cornell University collaboration employed the D.E. Shaw Research Anton 2 system hosted at PSC to simulate large molecules for long time periods. This capability shed light on how the TMEM16 scramblase proteins reorganize the cell membrane, a process critical to programmed cell death, blood coagulation and membrane repair.
Two articles reflect more traditional high performance computing successes. Georgia Institute of Technology researchers used Bridges to illuminate the physical chemistry of the two major pathways of electron transport. Bridges also helped a team at the University of Pittsburgh design a metal-oxide framework material that can turn effluent carbon dioxide into an industrially useful feedstock.
Since our last issue, our award-winning STEM education efforts have offered two more Microsoft-funded GCode—“Girls Code”—workshops to expose young women to potential careers in programming and computer science. Our Wide-Area Classroom of HPC workshops for researchers and computing professionals served 571 in-person and remote students between December 2018 and January 2019, putting us on track to exceed the previous year’s total of 3,015.
Finally, we’re proud to note that at the SC18 high performance computing conference in Dallas in November, PSC received a best-ever six HPCwire Awards for accomplishments in the field. You can find the details in this issue.