Prediction of reactivity in complex settings using machine learning

Machine learning (ML) has the power to revolutionize chemical synthesis. However, its adoption by the organic chemistry community has been limited due to the reliance on highly specific datasets for model development and the challenge of evaluating the trustworthiness of predictions. The overall goal of this project is to integrate, high throughput experimentation (HTE), ML, and reaction descriptions encoding mechanisms to build predictive models that not only accurately determine the feasibility of a reaction for complex substrates but also provide a measure of uncertainty associated with the prediction, enabling more reliable and informed decision-making in chemical synthesis. This approach will yield a deep understanding of the use case reactions and enable the rational adaptation of reaction conditions to different substrates. We hope that the models developed on specific reactions can be adapted to many cases and will accelerate synthesis planning and reaction development. Successful ML models require a good coverage of the chemical space explored, a numeric representation accurately describing the underlying chemistry, and an adapted ML architecture. My research will focus on (a) developing and benchmarking optimal chemical space sampling using active learning, (b) setup new workflows to quantify reaction outcome in a HTE fashion and (c) develop unprecedented dynamic graph-based reaction representations. These representations should pave the way toward transfer learning between reaction classes. This unmet challenge has the potential to unlock the power of ML for organic chemistry. The K99/R00 Pathway to Independence Award will allow me to continue to leverage the expertise of Prof. Reisman and the resources at Caltech Center for Catalysis to master HTE and overcome real-life obstacles encountered during synthesis by designing ML to answer these experimental issues. Caltech’s close-knit community and emphasis on interdisciplinary work will allow me to leverage expertise across different fields. Prof. Reisman’s collaborations with Prof. Yue and Prof. Chawla, both experts in computer science, will provide me with the support and guidance to further develop my skills in active learning. The mentoring of Prof. Coley will complement my training in ML for chemistry, specifically graph based representations. Prof. Alexanian will advise me on C–H functionalization. This support committee constituted of organic chemistry and computer science experts will provide the best guidance in shaping my research trajectory, identifying research interests, and building strong partnerships. I will look for conferences to perfect my communication skills and attend Caltech-sponsored trainings in career development, grantsmanship, responsible research, and teaching. My mentor and my collaborators will aid me in establishing a solid foundation for my career and ensure that my research and training align with my long-term goals. My ultimate career goal is to become an independent researcher and leader in the field of ML for chemistry, specifically focusing on developing organic reactions modeling tools that change how chemists approach synthesis. Project Number: 1K99GM163026-01 | Fiscal Year: 2026 | NIH Institute/Center: National Institute of General Medical Sciences (NIGMS) | Principal Investigator: Jules Schleinitz | Institution: CALIFORNIA INSTITUTE OF TECHNOLOGY, PASADENA, CA | Award Amount: $125,000 | Activity Code: K99 | Study Section: Special Emphasis Panel[ZRG1 CDB-Z (80)] View on NIH RePORTER: https://reporter.nih.gov/project-details/11283802