dr Marian Bubak
Sano Centre for Computational Medicine and ACC Cyfronet AGH, Krakow, Poland
websites: https://sano.science/, https://dice.cyfronet.pl/
Towards European Infrastructure for In Silico Medicine
Investigations in the area of in silico (computational) medicine [1], accelerated by the Physiome Project [2], are conducted by many researchers supported by the VPH Institute (currently the Society for In Silico Medicine), and outcomes are presented at the VPH Conferences [3]. The goal of this lecture is to present the emerging European infrastructure supporting research in computational medicine and the practical use of its results.
In 2022-2025, the consortium of the EU Project EDITH pursued a systematic approach to computational medicine, aiming to facilitate the implementation of the concept of digital twins in healthcare (DTH). It led to the idea of a virtual human twin (VHT) as an integrated environment enabling multi-scale, multi-temporal, and multi-disciplinary representation of quantitative human physiology and pathology [4]. The EDITH Roadmap [5] presents recommendations to enable efficient implementation of VHT related to assessment of creators and consumers, basic building blocks of VHT technologies and infrastructures, ELSI, standards, regulatory aspects, and sustainability conditions. With VHT one may: access every digital twin developed so far; search catalogue by data type, anatomical location, age of the patient; match any digital twin with any available dataset valid as input for that model; run models on every digital dataset available; orchestrate multiple digital twins to build multiscale or multisystem models; script it, and save scripts for automation or reuse. The Cyfronet and Sano Centre team developed a demonstrator to execute DTHs on HPC resources for scale-out studies, thereby enabling reproducible computer simulations.
The European Commission decided to support the practical implementation of the EDITH Roadmap as the European VHT initiative [6]. The first step is elaboration and operation of the VHT platform by the VHT NET Consortium (CINECA, Sano Centre, ENG, NTT Data, In Silico Trials). The demonstrator will be exploited in the development of the VHT platform
References
1. Peter Coveney, Roger Highfield: Virtual You. How Building Your Digital Twin will Revolutionize Medicine and Change Your Life, Princeton University Press, 2023
2. Physiom Project - https://www.auckland.ac.nz/en/abi/our-research/research-groups-themes/physiome-project.html
3. VPH Society for In Silico Medicine - https://vph-society.org/
4. M. Viceconti, M. De Vos, S. Mellone, L Geris: Position paper: from the digital twins in healthcare to the virtual human twin: a moon-shot project for digital health research, IEEE Journal of Biomedical and Health Informatics 28 (1), 491-501, 2023
5. EDITH European Virtual Human Twin https://www.edith-csa.eu/
European VHT initiative https://www.virtualhumantwins.eu/
doc. Mgr. Vít Vondrák, Ph.D.
IT4Innovations National Supercomputing Centre, VSB-Technical University of Ostrava, Ostrava, Czech Republic
website: https://www.vsb.cz/personCards/personCard.jsp?lang=en&person=von15
Current activities and further development of the national supercomputing infrastructure of the Czech Republic.
IT4Innovations National Supercomputing Center at the VSB – Technical University of Ostrava operates and provides the most powerful computing capacities for the research environment of the Czech Republic. Under the leadership of CESNET, the operator of the national research and education network, and together with CERIT-SC at Masaryk University, we form the strategic national research e-infrastructure, e-INFRA CZ. In addition to providing services of the e-INFRA CZ supercomputing infrastructure, IT4Innovations also conducts research activities in the areas of high-performance computing (HPC), large-scale data analytics, artificial intelligence, and quantum computing.
Currently, IT4Innovations operates the EuroHPC supercomputer Karolina, which is being integrated with the VLQ quantum computer acquired by the LUMI-Q consortium and EuroHPC. This integration is based on the LEXIS software platform developed at IT4Innovations, which enables simple, seamless, and secure access to diverse computing and data resources, including quantum computers. In addition, it enables the execution of complex computational workflows across different computing and data resources, including hybrid HPC and quantum computations. The LEXIS platform will also become a component of the European federated platform, integrating all EuroHPC computing resources.
The presentation will introduce the current and planned capacities of IT4Innovations, the policies governing access to them, and their use by user communities. It will also present the current services and the in-house-developed software platforms that extend e-INFRA CZ's capabilities. Finally, the newly established Czech AI Factory, including its planned services and potential collaboration topics with other AI factories, will also be introduced.
dr Olav Zimmermann
Jülich Supercomputing Centre (JSC), Jülich, Germany
website: https://www.fz-juelich.de/profile/zimmermann_olav
The stubborn old protein folding problem in the era of ExaFlops, Qbits, and AI
When, in 2024, the Nobel Prize in Chemistry was awarded to Demis Hassabis and John Jumper from Google DeepMind (developers of Alphafold2), many, including many scientists, claimed that the 50-year-old protein folding problem had been solved.
This talk will point out the difference between the structure prediction problem that Alphafold attempts to solve and the protein folding problem and explain why the latter is much harder and, apart from a few successes in the last decades, still essentially unsolved.
After presenting several semi-successful, pseudo-successful, and failed attempts of the community (including a large part of our own work using Monte Carlo simulations), the talk will try to dissect this old and stubborn problem and investigate the question of how new capabilities available at Supercomputer centers, such as exascale computing, quantum computing, and AI could be harnessed to finally solve it.
Along the way, I will also try to analyse why a single group at DeepMind could outcompete the entire academic community in the structure prediction competition CASP within a few years, what lessons can be learned from this, and why the protein folding problem might be well-suited to assess the current capabilities and competitiveness of academic research.