ZOZNAM ÚSPEŠNÝCH PROJEKTOV VO VÝZVE 1/24:
Molecular magnets for quantum technologies
AUTOR: Erik Čižmár
The project’s primary objective is to understand the influence of molecular design on magnetic anisotropy and exchange interactions, which affect quantum coherence and quantum entanglement in molecular magnets. The knowledge of the mechanisms of quantum coherence and quantum entanglement in molecular magnets is crucial for the promising application of molecular magnets in quantum information processing and quantum sensing. Nowadays, quantum chemical approaches, such as multireference ab initio methods as implemented in ORCA or MOLCAS packages, allow us to predict and study the properties of molecular magnets. Unfortunately, for large molecules with more than 120 atoms, the computational costs exceed the possibilities of small computational workstations. Within this project, we propose to study the influence of molecular structure on the magnetic anisotropy of a series of tetranuclear molecular spin clusters consisting of transition metal ions and/or lanthanide ions, which are suitable for further experimental and theoretical study of quantum coherence and quantum entanglement, using a multireference ab initio approach (SA-CASSCF/NEVPT2 or SA-CASSCF/DLPNO-NEVPT2) as implemented in the quantum chemical package ORCA.
Glycoside hydrolases of GH16, GH28 and GH72 families –their structure and protein – ligand interactions
AUTOR: Stanislav Kozmoň
Glycoside hydrolases of GH16, GH28 and GH72 families play a crucial role in the rearrangement of polysaccharide cell walls during physiological processes in plants and fungi. The complex study of these enzymes includes a number of experimental and computational approaches. The theoretical study is focused on the substrate specificity of these enzymes as well as potential inhibitors of proteins active on fungal cell walls with the aim to found active substances of potential future drugs. Part of this project focuses on the study of substrate specificity of plant xyloglucan endotransglycosylases and exopolygalacturonases utilizing molecular dynamics simulation as well as QM/MM study of their reaction mechanism and the other part focuses on design of possible inhibitors of fungi Crh and Phr enzymes that play a crucial role in virulence.
Simulation of smoke spread in roadway tunnel by FDS
AUTOR: Peter Weisenpacher
The proposed project focuses on the research of smoke propagation in roadway tunnels by Fire Dynamics Simulator (FDS). It builds on the results achieved in the previous research, which include measurements of airflow velocity and optical density in a real roadway tunnels as well as their computer simulations. Smoke tests performed in a roadway tunnel will be examined by FDS computer simulation and compared against the test measurements. The effect of computational grid resolution on the accuracy of the simulation and the influence of the decomposition of computational domain into meshes on simulation wall clock time and accuracy will be examined. The results of the project could contribute to the fire safety of tunnels.
Insights into the preaggregation conformation of intrinsically disordered protein tau
AUTOR: Ondrej Cehlár
As a member of the class of intrinsically disordered proteins (IDPs), tau protein in physiological state doesn’t acquire a clearly defined 3D structure, which makes the process of finding potential inhibitor of tau aggregation radically difficult. Therefore we aim to simulate molecular dynamics of truncated tau proteins capable to attain pre-aggregation tau conformation in monomeric state as shown by interaction with specific antibodies.
PRUDENCE – PRecise thUnDErstorm forecast in a chaNging ClimatE
AUTOR: Juraj Bartok
Spatially and temporally precise thunderstorm forecasts indisputably enhance the level of preparedness for natural hazards such as flash floods or tornadoes. Tourism and different types of transport, especially the aviation also benefit from more reliable thunderstorm forecasts. Moreover, the issue of thunderstorm-related natural hazards is getting even more relevant in the light of the extremization of weather phenomena due to the anthropogenic-driven climate change.
The PRUDENCE project, therefore, aims at addressing these problems by significant and measurable increase of the performance of thunderstorm forecast through the newest and promising artificial intelligence techniques. The basic pillars of PRUDENCE constitute of proven nowcasting approaches, making use of radar reflectivity data processed by the so-called optical flow concept (TITAN, TREC, RainyMotion, PySteps). These approaches and deep learning on satellite imagery, will be combined in the framework of ensemble stacking, which is a class of machine learning methods that uses outputs of several different methods as an input. Further improvement is expected with the use of additional meteorological-geographical inputs (lightning, rainfall intensities, topography, outputs of numerical forecast models).
The proposed methodology aims to construct the best nowcast in first hours horizon, with two major goals. First, to increase the probability of successful prediction of as many thunderstorms as possible, and secondly, to reduce the number of false alarms, which contributes to saving costs of preventive measures and, for instance in aviation, to better use the sector’s capacity.
Prediction of thermodynamically stable crystal structures of novel nickel oxides with the use of Evolutionary Algorithms and Density Functional Theory
AUTOR: Radovan Bujdák
Nickel oxides are attractive research targets due to broad spectrum of potential technological
applications including energy storage, (photo)catalysis, superconductivity, or transport conductive
oxides. However, the Ni-O binary system has not been sufficiently explored, and many of the
assumed nickel oxides remain either poorly characterized (e.g. NiO2 ) or completely unknown (e.g.
Ni2O, Ni2O5 ). In this project, we focus on ab initio computational modelling of these Ni-O phases
using state-of-the-art quantum-mechanical methods based on Density Functional Theory (DFT) and
Evolutionary Algorithms (EA). The aim of the project is to predict crystal structures and structure-
related properties including thermodynamical stability, electronic structure, magnetic ordering, and
lattice dynamics of the poorly characterized and the yet-unknown binary nickel oxide phases. The
outcome of the project will significantly broaden our knowledge of technological very important
nickel oxides and open new possibilities in materials science. Importantly, it will provide extremely
valuable source of information for experimentalists in synthesis and engineering of novel functional
materials with nickel. This project is a part of the Early-Stage Researcher grant of Faculty of
Materials Science and Technology of the Slovak University of Technology in Bratislava.
Prediction of new binary palladium oxides using evolutionary algorithms and Density Functional Theory
AUTOR: Diana Fabušová
This project focuses on computational modeling of binary palladium oxides using evolutionary algorithms (EA) and quantum-mechanical Density Functional Theory (DFT) methods. Its main goal is to predict thermodynamically stable crystal structures for selected stoichiometries of palladium oxides, such as Pd2O, Pd2O5, and PdO3. Palladium is a significant catalyst in many reactions with various technological applications. In addition to the pure metal, catalytic activity is increasingly attributed to palladium oxides, which are not yet sufficiently explored. The only thoroughly studied and technologically exploited palladium oxide phase is PdO. Research into new palladium oxides is therefore extremely important. The implementation of this project significantly contributes to expanding our knowledge of the [Pd-O] binary system. We will gain detailed information about the potential of palladium to form new binary phases with oxygen, their crystal chemistry, structure-property relationships, and the conditions of thermodynamic stability. Results of this study will ultimately lead to further opportunities for the technological use of palladium. This project is part of the “Early stage grant” (ESG) scheme (project no. 23-06-03-A) of the Slovak University of Technology.
Computational Infrastructure for the TERAIS project
AUTOR: Viktor Kocur
The goal of this project is to provide the researchers from the Department of Applied Informatics, Faculty of Mathematics, Physics and Informatics, Comenius Univerisity involved in the Twinning Horizon Europe project titled TERAIS with computational resources to carry out the project research. The main focus of the TERAIS project is human-robot interaction. This interdisciplinary research topic ties together various research fields such as robotics, cognitive science, psychology, deep learning, computer vision and natural language processing. The goal of the TERAIS project is to research ways in which humanoid robots can be made more trustworthy, human-aware and explainable. To pursue these goals, various computationally intensive methods can be employed do develop elements of the robotic system. Such systems may deal with the perceptive capabilities of the robot (computer vision, natural language processing), its expressive abilities (motion control, generation of descriptions) and its reasoning capabilities. In these areas, the latest trend is to use deep neural networks. Training, and other tasks utilizing these networks are computationally intensive, requiring high-performance GPUs. Training of such networks also requires large amounts of data, which needs to be processed or synthetically generated.
Influence of defects and disorder on power conversion efficiency of hybrid perovskite structures – DFT investigation
AUTOR: Kamil Tokár
Hybrid perovskites (PV) as optical absorbers in photovoltaic cells could possess a comparable theoretical photoelectric efficiency with advanced silicon technologies, while the technology for preparing such solar cells based on PVs should be much more advantageous. However, the efficiency of light conversion to photocurrent can be affected by the occurrence of various defects states in the structure of the PV film and also by the thermal dependence of optical absorption.
The first aim of the project will be to study the influence of basic types of defects-vacancies on the electronic band structure of MAPbI3,(Br3,Cl3), dielectric functions and optical absorption transitions using the quantum Density Functional Theory (DFT and hybrid-DFT) methods.
In the next part, the project will focus on the description of the fine structure of the levels near the conduction band and the impact of thermal lattice vibrations induced by phonon modes on the PV optical absorption edge. In the study of lattice dynamics, it is supposed to simulate lattice dynamics using DFT and electronic levels lying close to the PV conduction band using post-DFT methods (TD-DFT, G0W0, BSE).
Simulations of key electronic and photoelectric parameters will provide modelling support to the experimental part focusing on the passivation of defects during PV films growth and tuning the opto-electronic characteristics.
Lean Neural Networks
AUTOR: Vladimír Boža
We plan to develop methods for creating neural networks which require less computational resources. Our primary focus is the pruning of neural networks. We will apply our newly found techniques in general benchmarks of neural networks and also in bioinformatics and biomedicine to improve the accessibility of state-of-the-art tools.
Generative AI for Medical Image Processing
AUTOR: Peter Drotar
This project aims to harness the capabilities of generative AI models for advancing medical image processing, focusing on critical aspects such as image out-painting and speckle removal in ultrasound imaging.
Generative models, encompassing Generative Adversarial Networks (GANs) and Diffusion Models, are instrumental in various AI applications. GANs, introduced by Goodfellow and colleagues, engage in adversarial training to create realistic data, while Diffusion Models offer a principled approach to image denoising through a gradual noise removal process.
In the context of medical imaging, the project’s primary objectives are twofold. Firstly, it seeks to employ GANs for medical image out-painting, specifically expanding the field of view in cardiac ultrasound. The competitive training dynamics of GANs will be utilized to generate realistic and contextually coherent outpaintings, aiding in obtaining necessary views for accurate diagnosis.
Secondly, the project aims to address the challenge of speckle noise in ultrasound images. Stable diffusion models will be designed and implemented to effectively denoise ultrasound images. The parameterization and training of these models on noisy-clean image pairs will ensure optimal denoising parameters, preserving essential image features crucial for accurate diagnostics.
To meet the computational demands of training GANs and stable diffusion models, the project emphasizes the utilization of GPUs within High-Performance Computing (HPC) environments. Efficient parallel processing will expedite training, enhance model convergence, and facilitate rapid experimentation with diverse model architectures and hyperparameters.
The anticipated outcomes include the publication of scientific papers on image out-painting and speckle removal in medical imaging. The project aims to contribute to the broader understanding of generative AI applications in medical image processing, ultimately leading to improved ultrasound image quality and enhanced diagnostic accuracy. The utilization of Devana’s GPU resources will play a pivotal role in achieving these outcomes, enabling efficient parallel processing and contributing to the overall success of the project.
Inclusion of conformational variability: a promising route for better predictions of binding affinity of inhibitors of SARS-CoV-2 main protease.
AUTOR: Michal Pitoňák
Although the SARS-CoV-2 virus (and COVID-19 disease) seems not to be the hottest topic nowadays anymore, the list approved treatments for COVID-19 still remains rather short [1]. Since the threat of emergence of new, possibly more dangerous mutation variants, or even other virus type cannot be completely neglected, developing pipelines for rapid discovery of new drugs is as important as ever. Introducing a novel drug to market is obviously a complicated, time- and money-consuming process that has plenty of stages. In this proposal we focus on rapid, AI/ML-based (Artificial Intelligence / Machine Learning) approach for reproducing so-called docking score obtained using AutoDock Vina [2], with a special focus on exploiting information regarding the conformational variability o ligands.
[1]: P. Shree, P. Mishra, C. Selvaraj, S. K. Singh, R. Chaube, N. Garg, Y. B. Tripathi, JBSD 2022, 40(1), 190.
[2]: J. Eberhardt, D. Santos-Martins, A. F. Tillack, S. Forli, J. Chem. Inf. Model. 2021, 61(8), 3891.
Advanced materials for green technological applications
AUTOR: Eva Scholtzová
The project’s primary scientific goal is to compare the adsorption properties of layered structures (LS) based on the graphene clays and their modifications by studying the interactions and stability of pollutant-LS complexes essential for eliminating pollutants from wastewater and soil. From the ecological point of view, the knowledge of the keying mechanism of contaminants on the surface of the LS has a crucial meaning. Structural stability, interaction energies, and mechanism of keying in the formation of complexes among the selected adsorbed pollutants will be studied for both the LS of the economically more affordable AS and the more expensive materials based on graphene by computational methods (e.g., DFT-D3 method, ab initio molecular dynamics).
Pioneering Benchmarks and Enhancing Reasoning Abilities of Large Language Models in Slovak – A Pathway to Advancing Low-Resource Language AI
AUTOR: Marek Šuppa
This project aims to advance AI research in low-resource languages by focusing on developing benchmarks and enhancing reasoning abilities of Large Language Models (LLMs) for Slovak. The goals include creating the first benchmarks for LLM evaluation in Slovak, fine-tuning pre-trained models for specific reasoning tasks, and improving LLM reasoning capabilities. To achieve this, the project will employ computational methods such as developing culturally relevant benchmarks and exploring various transformer-based architectures like LLaMA, Mistral, PolyLM for model adaptation. Access to substantial computing power is crucial for the training, fine-tuning, and evaluation processes. The project, leveraging the collaborative efforts of students from the Faculty of Mathematics, Physics and Informatics at Comenius University in Bratislava, is poised to make significant contributions to AI research in low-resource languages.
Luminescent Probes and Photosensitizers for Two-Photon Excited Bioimaging and Photodynamic Therapy
AUTOR: Peter Hrobárik
Molecules with significant two-photon absorption (TPA) serve as active components in modern laser technologies, such as microfabrication or high-capacity optical data storage. Extensive TPA cross-sections combined with excellent luminescence can be achieved by using small building blocks based on S,N-heteroaromatic compounds with a proper arrangement of electron-donating and/or electron-withdrawing substituents, and by complexation of these scaffolds featuring intramolecular charge transfer with selected metal ions. The most efficient derivatives and substitution patterns will be identified using computer-aided screening (rational design based on quantum-chemical calculations). The aim of the project is to functionalize our recently developed TPA-active scaffolds / dyes in order to obtain other essential properties, such as increased solubility in polar media, introduction of spin-orbit coupling allowing spin-forbidden singlet-triplet transition and production of highly reactive singlet dioxygen and/or prolongation of excited-state lifetime to the microsecond time domain (transition from fluorescence to phosphorescence). These attributes, in conjunction with TPA activity, will facilitate the use of the developed systems in practical applications, with a special focus on „multichannel“ nonlinear imaging of biological structures by high-resolution microscopy and photodynamic therapy using deep-penetrating near-infrared laser radiation.
The computational research will systematically benefit from using state-of-the-art quantum-chemical calculations used primarily in predicting linear and nonlinear optical properties and singlet-triplet transition efficiencies. QCH calculations will be performed using modern TD-DFT methods with local hybrids and long-range corrected functionals as well as with more robust coupled-cluster methods, such RI-CC2, RI-ADC(2) and RIADC(3), which include necessary dynamical and static electron correlation effects to describe charge-transfer processes more accurately.
Ab initio modelling of novel binary silver halides
AUTOR: Mariana Derzsi
This project focuses on ab initio modelling of binary silver fluorides and chlorides using Density Functional Theory (DFT) methods and evolutionary algorithms (EA) for crystal structure prediction. Silver in conjunction with halogens has the potential to create materials with exceptional properties, which may make them the multifunctional materials of the future. This potential is being intensively explored by us in cooperation with partner organizations by means of quantum-mechanical modelling and experimental approaches. The Ag-F system exhibits richness of unique phases, which account for an important cuprite analogue, a potentially first silver fluoride multiferroic, a new unique form of nanotube, and several charge-ordered phases. This diversity contrasts with only one known phase in the binary Ag-Cl system. Our recent DFT and EA calculations have suggested favorable stability of various AgClx (x=0.5–4) phases with unexpectedly rich crystal chemistry, while some of our results are already getting support also from the most recent experiments. The aim of the current project is to characterize the newly discovered phases, explore possible means of their functionalization and uncover other new phases in the Ag-H systems (where H is halogen) using EA and DFT methods. The outcomes of the project will significantly broaden our knowledge of crystal chemistry and physics of silver halides.