Radim Špetlík and Jan Hlavsa from the Faculty of Electrical Engineering of the Czech Technical University in Prague (CTU), in collaboration with Petra Pojmanová and Jana Čechová from the University of Chemistry and Technology in Prague (UCT), use the LUMI supercomputer to enhance the analysis of human scent using machine learning. They are now trying to identify distinct scent profiles, which could also be beneficial in other areas, such as forensic investigations.

Human scent carries valuable information about our health, behaviour, and identity. Analytical methods such as two-dimensional gas chromatography complemented by mass spectrometry can separate and identify these complex chemical compounds. However, the challenge remains in post-processing the large volumes of data that these methods generate.

This is precisely where advanced machine learning, specifically deep learning and computer vision techniques, come into play, enabling such large volumes of data to be analysed efficiently. “Our main goal is to develop advanced methods for analysing data obtained by analytical methods,” says Radim Špetlík. “Using convolutional neural networks (CNNs), we can extract useful information directly from the raw data and verify identities with minimal manual intervention,” he explains. “This research could fundamentally contribute to new methods not only in forensics but also in medicine, such as the early detection of diseases based on scent,” he adds.

Access to powerful computing resources is essential for analysing complex and large-scale data. Therefore, Radim Špetlík uses not only the Karolina supercomputer but also LUMI, which is one of the most powerful supercomputers in Europe and is available to Czech scientists through the IT4Innovations Open Access Grant Competitions.

Radim Špetlík has obtained additional computational resources of the LUMI supercomputer, particularly its GPU partition, which can be used until January 2026. “The research continuation is crucial for us, so I have requested access to the LUMI supercomputer again. With LUMI, I can take my research into automating and refining the analysis of complex human scent profiles to a new level,” concludes Špetlík.

 

*Photographs from the research presentation at the Multidimensional Chromatography Workshop 2025