This paper introduces AiZynthFinder, an open-source software for retrosynthetic planning based on Monte Carlo tree search and a neural network policy. The software is designed for speed and flexibility and has been shown to generate plausible synthetic routes, although comprehensive benchmarking and experimental validation are areas for future development.

The authors disclose affiliations with AstraZeneca and the University of Bern. While the software itself is open-source, the potential for bias in choosing comparisons or emphasizing certain aspects cannot be ruled out.

This paper presents a valuable open-source tool for retrosynthetic planning. While the core algorithm shows promise and the open-source nature fosters community development, the limited comparison with other tools and the reliance on commercial data prevent a full 5-star rating.