Next-Gen HPC: The Path to Exascale – Artifical Intelligence and Personalized Medicine – an event report

Internationally renowned researchers from the HPC and personalized medicine community converged at the Leibniz Supercomputing Centre (Leibniz-Rechenzentrum, short LRZ) in Garching, Germany on June 29, 2018 to offer their expertise and insight in a symposium titled 'Next-Gen HPC: The Path to Exascale - Artificial Intelligence and Personalized Medicine’.

Keynote speakers were Dr. Alan Gara, Chief Architect at Intel with his presentation ‘Technology Challenges and Trends – a Look through a 2030 Crystal Ball’ and Dr. Fred Streitz, Director of the High Performance Computing Innovation Center at the Lawrence Livermore National Lab speaking on ‘Machine Learning and Predictive Simulations – HPC and the US Cancer Moonshot’.

Distinguished speakers included: Prof. Satoshi Matsuoka, Director of the Riken Center for Computational Science (Japan); Prof. Rick Stevens, Deputy Director of the Argonne National Laboratory and Professor at the University of Chicago; Prof. Peter Coveney, Director of the Centre for Computational Science at University College London (UK) and Dr. Barbara Schormair, Deputy Director of the Institute of Neurogenomics at the Helmholtz Zentrum München, German Research Centre for Environmental Health. The panel was moderated by Dr. Thomas Friese, Siemens Healthineers and Prof. Martin Schulz, Technical University of Munich.

The speakers provided detailed insights into trends and challenges, showing how closely the three topics of HPC, artificial intelligence and personalized medicine are interlinked. All speakers agreed that future supercomputers and their greatly increased computing capacity will enable new interactions of computer simulations, machine learning and the analysis of large amounts of data. However, as pointed out by Dr. Gara, the HPC community has to solve some challenges along the way, such as fundamental improvements in data transport and storage, changes in system architectures to scale memory, compute and communications or power consumption. Prof. Matsuoka added that the big data and ML communities must bring some of the HPC rigor in architectural, algorithmic, system software performance and modeling into their fields to make the convergence of HPC and ML fully realized.

In his keynote, Dr. Fred Streitz, one of the lead researchers in the Cancer Moonshot Project of the US Department of Energy (DOE), assured that increasing the compute capacity of supercomputers from the present petaflop to the future exaflop fundamentally changes the scientific questions that can be asked and, above all, answered. Pointing to his team’s work on machine learning guided multiscale simulations for investigating Ras cancer proteins, such increased computing amplifies the abilities of machine learning in predictive simulations. Prof. Coveney (UCL) agreed, noting the fledgling use of predictive simulation in the field of medical and pharmaceutical research, and highlighting the impact possible through presentation of his work using predictive simulations to determine which breast cancer treatment is best suited to treat patients based on their individual genetic profile.

Dr. Barbara Schormair, an expert in neurogenetics, asserted the necessity of an interdisciplinary approach bringing together technicians, lab scientists, engineers, computational scientists, biologists, neurologists and doctors to increase scientific knowledge at the crossroads of personalized medicine, HPC and machine learning. She noted the USA, Great Britain and the Netherlands have already made considerable progress with major projects in the field of human genomics and presented concrete examples of how patients benefit directly from current research at the Helmholtz Zentrum München and the University Hospital rechts der Isar at the Technical University of Munich (TUM). For example, advances in genetics such as exom sequencing allow more precise diagnoses, e.g. in dystonia, and in certain cases even the use of tailor-made therapies.

In the subsequent panel discussion, lively audience discussion with the experts focused on questions around ethical handling of patient data and on the possibilities now available through the new technologies. The experts also intensively discussed how early career scientists can become educated in cross disciplinary fields and how interdisciplinary cooperation between experts in supercomputing and ML as well as physicians and geneticists can be promoted and successfully shaped.

The Next-Gen HPC event attracted representatives from industry, high-ranking officials of the Bavarian Ministry of Science and the Arts, as well as professors and PhD students from Germany’s leading universities of excellence, the Ludwig-Maximilians-Universität, Munich (LMU) and the Technical University of Munich (TUM).

"We are very pleased to bring internationally renowned speakers to LRZ for this discussion. In the presentations and the panel, it became clear how well positioned Bavaria is for next-generation supercomputing and its application to artificial intelligence and personalized medicine. To continue pushing forward and to achieve the breakthroughs we know are possible, building and fostering rigorous interdisciplinary exchange and international cooperation is paramount to our common success and future computing," said Prof. Dieter Kranzlmüller, director of the LRZ.