Automation and AI in Molecular Materials Discovery
Abstract – In my talk, I will focus on our first results towards establishing a high-throughput synthesis and analysis platform for peptides and peptide derivatives. This platform enables us to perform data-driven active learning for peptide design, in which we combine experiments with machine learning and AI. We have been using this platform for both fundamental and more applied investigations, and I will show some results on analysis of fundamental molecular properties, such as solubility, and more applied properties, such as interfacial tension. Lastly, I will show some of our recent efforts involving infusion of prior literature into our active learning decisions through agentic AI.
Bio – Mathijs received his BSc. and MSc. in chemistry from the Radboud University, before moving to the Eindhoven University of Technology for a PhD in the group of Bert Meijer. Here, he worked on model-driven understanding of competitive interactions of solvents and additives in supramolecular polymerizations. In 2021, Mathijs changed fields and joined the lab of Jörn Piel at the Intsitute of Microbiology at ETH Zurich. During his time at ETH, Mathijs developed data-driven workflows for natural product discovery and engineering of polyketide synthases. In 2024, Mathijs joined the Radboud University and Big Chemistry consortium as a group leader. Merging his prior experience in supramolecular chemistry and protein engineering, his group aims to develop data-driven methods to for peptide material design, combining high-throughput experimentation with machine learning, for applications in biomedicine and materials science.