Single-step Retrosynthesis via Reaction Center and Leaving Groups Prediction

Retrosynthesis aims to break down desired molecules into accessible building blocks in a systematic manner. However, current template-based retrosynthesis approaches face limitations due to a fixed set of training templates, hindering their ability to discover new chemical reactions. To overcome this challenge, we present a novel retrosynthesis prediction framework that can generate new chemical templates beyond the training set. This innovative approach has demonstrated superior performance compared to previous template-based methods. Additionally, we propose a method to improve atomic mapping accuracy by matching the maximum common subgraph of the reaction center and leaving group in the retrosynthetic template. This enhancement effectively increases the precision of atomic mapping settings.