Photocatalytic Substrate Engineering through Ensemble Deep Learning

Objective

Over the past two decades, visible-light photocatalysis has revolutionized greener and more efficient chemical synthesis, significantly impacting pharmaceuticals, agrochemicals, and materials. However, a key challenge remains: the need to discover new substrates that can reliably and sustainably harness light energy. While current research has improved the prediction of light-driven reactions, existing methods are often slow, costly, and rely heavily on trial-and-error experimentation. The PROJECT, PHOTO-SEED, aims to pioneer an interdisciplinary approach, including Chemistry, Computer/Data Science, and Chemical Informatics. The project will: (a) create the first high-quality database of light-sensitive molecular properties, (b) develop deep learning models to provide accurate and reliable predictions across a wide range of molecules, (c) use generative AI to design synthetically feasible candidates, and (d) validate these discoveries experimentally, particularly in reactions relevant to medicinal chemistry. By integrating artificial intelligence with chemistry, PHOTO-SEED will significantly reduce discovery cycles from months to hours, minimize environmental impact through predictive design, and unlock new opportunities for sustainable synthesis. The urgency of this project directly supports Horizon Europe’s objectives of advancing green technologies and digital transformation. The fellowship offered through this project will be transformative for the researcher, providing advanced expertise at the intersection of AI and catalysis, opportunities for international collaborations, and the development of leadership skills. The host, Prof. Frank Glorius, an internationally recognized expert with a successful track record of supervising MSCA fellows, will provide the necessary resources and mentorship to ensure the success and impactful outcomes of this initiative.