Computer-aided retrosynthetic design: fundamentals, tools, and outlook

Identifying synthetic routes to molecules of interest has been one of the most challenging tasks for synthetic chemists. We have witnessed the gradual adoption of computational tools in solving retrosynthetic design problems for the past 50 years. Especially in the past five years, computer-aided retrosynthesis publications have become more common due to advancements in computing power, data availability, and data driven algorithms. This paper provides a review of contemporary retrosynthesis methodologies. We define the retrosynthesis framework and describe how machine learning techniques contribute to reaction template extraction and synthetic complexity ranking. We explore template-based and template-free synthetic search strategies and discuss how learning algorithms can prioritize the most applicable transformation rules. We conclude by addressing potential challenges and opportunities facing automated synthetic planning.