Users and their projects

All-Atom Any-Modality Molecular Diffusion Model for Biomolecule Design

Call: 34th Open Access Grant Competition; OPEN-34-1

Researcher: Anton Bushuiev

Institution: Czech Technical University in Prague, CIIRC

Field: Biosciences

Researchers at CIIRC CTU will harness the power of one of Europe’s most powerful supercomputers, LUMI, to develop artificial intelligence that assists in the design of new medicines. This advanced model learns from three-dimensional images of molecules how they interact – or, conversely, compete – with one another. Using these insights, it can propose new compounds that precisely target molecules associated with diseases such as viral infections or cancer. Scientists will then test these designs in the laboratory to verify their effectiveness. The resulting tool will also be made available to the wider scientific and medical communities.

This research is part of the CLARA Centre of Excellence, funded by the Johannes Amos Comenius Programme.

Generative models for bone dataset augmentation

Call: 34th Open Access Grant Competition; OPEN-34-72

Researcher: Martin Špetlík

Institution: Technical University of Liberec

Field: Biosciences

A research team from the Technical University of Liberec is using the LUMI supercomputer to create an extended dataset of clinical CT bone scans through generative machine learning models. The aim is to gain a deeper understanding of how differences in bone shape and structure between individuals relate to bone remodelling – a process essential for healthy bone function.

The scientists are investigating whether CT scans can reveal the degree of mechanical loading experienced by patients’ bones. Such insights could pave the way for more precise and individually targeted treatment of osteoporosis. To address the shortage of available clinical data, they are generating realistic synthetic CT scans using advanced generative neural networks.

The findings could contribute to earlier diagnosis, more effective treatment and fracture prevention, thereby reducing both the health and economic burden of an ageing population.

This research is part of the Czech Science Foundation project GA ČR 24-10862S – Data-driven modelling of bone morphology and mineral density.

Machine learning-based surrogate modeling for conjugate heat transfer enhancement with complex surface geometry: CFD simulation

Call: 34th Open Access Grant Competition; OPEN-34-86

Researcher: Jan Boháček

Institution: Brno University of Technology

Field: Engineering

Jan Boháček from Brno University of Technology will utilise the supercomputers Barbora and Karolina to develop a computationally efficient and accurate framework for modelling conjugate heat transfer in complex surface geometries. These calculations are typically performed using CFD simulations, which, at high accuracy, demand substantial computational resources.

The aim of the project is to employ machine learning methods to create so-called surrogate models that can replace parts of these resource-intensive simulations. This approach will reduce computational costs while maintaining result reliability, enabling more efficient optimisation of heat transfer.

The research is supported by the Czech Science Foundation project GA ČR 25-16935S.

Properties of liquids interacting with pressurized gases: from natural gas to hydrogen economy

Call: 34th Open Access Grant Competition; OPEN-34-13

Researcher: Jan Heyda

Institution: University of Chemistry and Technology Prague

Field: Material sciences

The dissolution of gaseous methane (under high pressure of 105 bar) and its diffusion in liquid p-xylene take place inside a 9 mm titanium measuring cell. The temporal evolution of these processes, including the position and shape of the interface (green and pink curves), is monitored using a unique neutron imaging method. Molecular dynamics simulations provide a sub-nanometre view of the arrangement of methane in the gas phase (black spheres), at the interface (blue), and within (yellow) the liquid p-xylene (turquoise region). By combining experiments and simulations, it becomes possible to evaluate key physico-chemical properties of these systems under conditions closely resembling those encountered in the energy industry.

Scientists from the University of Chemistry and Technology in Prague will harness the power of the Barbora and Karolina supercomputers to investigate how selected liquids interact with gases such as methane, ethane and hydrogen under high pressure. These insights are crucial both for today’s natural gas industry and for the emerging hydrogen economy. In current energy systems, methanol serves as an inhibitor of undesirable methane hydrate formation.

Looking ahead, methanol and toluene are considered promising liquid organic hydrogen carriers (LOHCs) for future energy infrastructures. Using molecular simulations, the research team is examining non-bonded interactions in detail, thereby complementing unique neutron imaging of high-pressure systems carried out in collaboration with scientists from the Paul Scherrer Institute (PSI).

The research forms part of the Czech–Swiss project “Properties of Liquids Exposed to Compressed Methane, Ethane and Hydrogen” (GA ČR–SNSF 23-04741K).

Exploring the mechanisms of plasticity in nitinol

Call: 34th Open Access Grant Competition; OPEN-34-88

Researcher: Miroslav Černý

Institution: CEITEC

Field: Material Sciences

An atomistic simulation of plastic deformation in the martensitic structure of Nitinol. Within this structure, researchers have identified a new plastic deformation mechanism – known as kwinking (a term coined by combining twinning and kinking) – which merges the processes of twinning and anisotropic plastic slip.

A research team from CEITEC, Brno University of Technology and the Czech Academy of Sciences is harnessing the power of the Karolina, LUMI and Barbora supercomputers to study in detail the properties of a nickel–titanium alloy – Nitinol (NiTi). Thanks to its ability to return to its original shape after deformation and subsequent heating – a phenomenon known as shape memory – Nitinol is a unique material with a wide range of applications, from medical devices to technical components. In addition to its shape memory effect, Nitinol also displays a surprisingly high ductility, i.e. the capacity for plastic deformation.

This remarkable combination of properties stems from mechanisms at the atomic level that remain largely unexplored. To uncover them, scientists are employing novel approaches in atomic simulations based on machine learning and quantum-mechanical calculations, which now make it possible to reconstruct the behaviour of materials down to the level of individual atoms.

The research, carried out within the Czech Science Foundation project GA ČR 25-16285S, “Kwinking concept in NiTi shape memory alloy”, paves the way for the development of new materials that combine the unique features of shape memory alloys with the high toughness of materials utilising TRIP and TWIP mechanisms.