Data-Driven Decision Making for Reaction Discovery
Synthetic chemistry requires decision making based on numerous factors, such as reactants, reagents, reaction conditions, or procedures. Our lab aims to enhance this decision making process: We develop (Bayesian) machine learning tools for chemical reactivity, leveraging its power to find patterns in complex, high-dimensional data.
Our aim is to create ML tools that are directly useful in a synthetic organic chemistry lab. Importantly, these tools are meant to add to, not replace, the knowledge that expert chemists already have. Examples of such tools include:
- Machine learning for predicting the outcomes of new catalytic reactions
- Optimization algorithms to find and improve reaction conditions
- Data-driven strategies for exploring reaction mechanisms.
Our software tools are developed primarily in Python, making use of established libraries for machine learning and Bayesian optimization (e.g. PyTorch and BoTorch).
Digital Tools for the Synthetic Laboratory
“Digital transformation in […] labs is progressing. However, we also found that advancements with talent and technology aren’t happening at the needed pace.”
(Accenture Business Report, 2022)
Our research aims to streamline the way we generate, collect and analyze data in the lab. To do this, we make sure that we can digitally interface with existing laboratory equipment – and at the same time, we integrate new robotic tools (e.g. pipetting robots) into our workflows. Analytical instruments such as GC-MS or LC-MS are crucial for data generation. Therefore, systematically extracting, processing and analyzing this data is a key aspect of our work. Ultimately, our goal is to generate more usable information from every single experiment – to be able to make smarter, more informed decisions.
Discovery of Sustainable Catalytic Processes
With the (expected) change in resource availability, synthetic chemistry needs to undergo a transformation from relying on crude oil to using renewable feedstocks. This shift demands new approaches in organic chemistry – finding new precursors and platform chemicals, and developing new classes of reactivity.
Our vision is to discover these new reactions, and develop them in our lab using the toolbox described above. Therefore, we focus on catalysis, and particularly homogeneous catalysis with earth-abundant metals (e.g. iron), as a key technology: The unique radical reactivity of these metals can lead to new and unconventional methods for forming and breaking chemical bonds.
