Autonomous optimisation of general conditions for novel reaction

This proposal, AOGCNR, proposes to develop a fully autonomous, data-driven methodology for optimising general conditions for a novel organic catalysis reaction. It integrates a self-driving laboratory (SDL) with a Large Language Model (LLM)-enhanced Bayesian Optimisation framework, combining cutting-edge robotics, machine learning and expert chemical reasoning. A two-phase approach is proposed. The first phase uses a statistical pipeline to screen a vast chemical space of over 3 million combinations, reducing it by over 97% to a manageable subspace. The second phase deploys the LLM-enhanced Bayesian Optimisation algorithm within this refined space to autonomously identify general reaction conditions that perform well across multiple substrates. This approach represents a significant advancement over current methods, which typically focus on single-substrate optimisation. The methodology will be validated on both a known reaction and a novel cyclisation discovered by researchers at the University of Liverpool, demonstrating its ability to accelerate the discovery of novel catalytic reactions.This project's impact is significant on multiple levels. It advances chemical research by enabling data-efficient exploration of complex reaction spaces. Societally, it democratises access to advanced chemistry tools. Economically, this approach offers a scalable and cost-effective alternative to traditional trial-and-error methods, streamlining industrial research and development, and contributing to more sustainable and efficient innovation.