Investigating the Reliability and Interpretability of Machine Learning Frameworks for Chemical Retrosynthesis

Abstract

Machine learning models for chemical retrosynthesis have attracted substantial interest in recent years. Unaddressed challenges, particularly the absence of robust evaluation metrics for performance comparison, and the lack of black-box interpretability, obscure model limitations and impede progress in the field. We present an automated benchmarking pipeline designed for effective model performance comparisons. With an emphasis on user-friendly design, we aim to streamline accessibility and facilitate utilisation within the research community. Additionally, we suggest and perform a new interpretability study to uncover the degree of chemical understanding acquired by retrosynthesis models. Our results reveal that frameworks based on chemical reaction rules yield the most diverse, chemically valid, and feasible reactions, whereas purely data-driven frameworks suffer from unfeasible and invalid predictions. The interpretability study emphasises that incorporating reaction rules not only enhances model performance but also improves interpretability. For simple molecules, we demonstrate that Graph Neural Networks identify relevant functional groups within the product molecule, providing thermodynamic stabilisation over the reactant precursors. In contrast, the popular Transformer fails to identify such crucial stabilisation. As the molecule and reaction mechanism grow more complex, both data-driven models propose unfeasible disconnections without offering a chemical rationale. We stress the importance of incorporating chemically meaningful descriptors within deep-learning models. Our study provides valuable guidance for the future development of retrosynthesis frameworks.

Friedrich Hastedt

Supervised Machine Learning for Retrosynthesis and Synthesis Route Planning

Friedrich is a PhD candidate at the CDT React with a research focus on developing interpretable and reliable machine learning frameworks for chemical retrosynthesis. He is generally interested in the application of deep learning techniques for small molecules and proteins. Prior to his PhD, he was an undergraduate student in the Department of Chemical Engineering at Imperial College London.

Dr. Ehecatl Antonio del Rio Chanona

Principal Investigator of OptiML

Antonio del Rio Chanona is the head of the Optimisation and Machine Learning for Process Systems Engineering group based in thee Department of Chemical Engineering, as well as the Centre for Process Systems Engineering at Imperial College London. His work is at the forefront of integrating advanced computer algorithms from optimization, machine learning, and reinforcement learning into engineering systems, with a particular focus on bioprocess control, optimization, and scale-up. Dr. del Rio Chanona earned his PhD from the Department of Chemical Engineering and Biotechnology at the University of Cambridge, where his outstanding research earned him the prestigious Danckwerts-Pergamon award for the best PhD dissertation of 2017. He completed his undergraduate studies at the National Autonomous University of Mexico (UNAM), which laid the foundation for his expertise in engineering.