Researcher
Building C7.1
Room 0.013
Universität des Saarlandes (Saarland University)
66123 Saarbrücken
mrdupont (at) lsv dot uni-saarland dot de
Research Interests
I’m working on improving machine learning for molecular property prediction — specifically, I’m trying to understand what drives the gap between chemical language models and graph neural networks in this domain.
As part of the RTG Neuroexplicit Models, I’m exploring how physical priors like symmetries and conservation laws can be baked into neural network architectures. More broadly, I’m interested in what makes a good representation and why self-supervised methods tend to work so surprisingly well.
Publications
Sultan, A., Rausch-Dupont, M., Khan, S., Kalinina, O., Klakow, D., & Volkamer, A. (2025). Transformers for molecular property prediction: Domain adaptation efficiently improves performance. arXiv preprint arXiv:2503.03360.
Andersen, L., Rausch-Dupont, M., Martínez León, A., Volkamer, A., Hub, J. S., & Klakow, D. (2025). Accelerating ligand discovery by combining Bayesian optimization with MMGBSA-based binding affinity calculations. bioRxiv, 2025-06.
Students
If you are interested in writing a thesis with me, please email me about your research interests and include a transcript of records.