Supercomputing for Medicine and Drug Research
What happens in the human body during a heartbeat? We don't really know actually. Even if the functioning of the heart, brain and organs is known, we can - if at all - only take snapshots of their work. Computer technology helps to track down the human being: For a few years now, an international research team has been developing and researching software, simulation algorithms and codes for medicine and pharmacy. The goal of the CompBioMed Centre of Excellence is the digital twin for the representation of body functions. So IT was also the focus of the CompBioMed Conference in mid-September, in which the Leibniz Supercomputing Centre (LRZ) was also involved. Prof Dr Dieter Kranzlmüller, head of the LRZ, and Dr David Wifling from the ComputationalX team moderate the panel "Towards Exascale – Validation, Verification and Quantification of Uncertainties", Elisabeth Mayer from the Centre for Visualisation and Virtual Reality (V2C) showed the visualization of blood flow in the forearm.
Blood flow in the forearme. Graphics: LRZ
"Visualizations, models and simulations not only illustrate body functions such as blood flow, they now also give a better insight into molecular processes," Wifling summarises impressions from the conference. "That's why more and more computing power and computer performance is needed, hopes rest on exascale systems, but also on quantum computing." On the one hand, modelling down to the molecular level is going into more and more detail; on the other hand, researchers want to simulate body events for longer. The exciting simulation and visualisation of blood flow in the forearm illustrates this: it shows what happens during a heartbeat in the forearm. For the short sequence, more than 220 million data points were elaborately calculated, modelled and visualised. More than 7 gigabytes of data were collected per output step. Only supercomputers can handle such quantities, but they have their limits: Doctors and researchers could learn more from it if they could watch the blood flow for longer. At CompBioMed, they are working also on advancing simulations from the micro to the millisecond range and are also looking at the next performance levels of high-performance computing (hpc): "That’s really a difficult challenge, Exascale therefor is becoming more and more important for CompBioMed. The first systems that the centre of excellence wants to use will soon be set up in Europe," reports Wifling.
Researchers discussed during the Virtual Human Conference how existing CompBioMed software such as HemeLB (blood flow) or TIES (molecular compounds) can be adapted to new Graphic Processing Units (GPU) and to the new systems, thereby compensating for uncertainties in computing. Another topic: pattern recognition and machine learning as methods for diagnosis and the search for active agents against diseases. The centre of excellence has participated in the research around COVID-19, searching for substances against SARS-Cov-2 viruses and exposing its techniques to a practical test. In addition to active substances, ideas were collected for optimising smart systems that help improve drug research. SuperMUC-NG also searched through libraries of chemical substances for CompBioMed. Such tasks can be better organised with the QCG Pilotjob software. This is being tried out in practice at SuperMUC-NG: "With QCG Pilotjobs, several simulations on molecular dynamics can be distributed and processed simultaneously on the nodes of an HPC system," explains Wifling. Whether and how this helps to accelerate the search for effective substances will possibly be the topic of the next CompBioMed conference.
Read more about the challenges of CompBioMed, the partnership between the LRZ and the Centre of Excellence as well as background information on the digital twin and what it means for medicine and pharmacy on the LRZ website. (vs)