AiZynthFinder: a fast, robust and flexible open-source software for retrosynthetic planning

Area of Science:

• Computational Chemistry
• Cheminformatics
• Artificial Intelligence in Chemistry

Background:

• Retrosynthetic planning is crucial for chemical synthesis.
• Existing methods can be time-consuming and complex.
• Need for efficient and accessible tools for retrosynthesis.

Purpose of the Study:

• Introduce AiZynthFinder, an open-source software for retrosynthetic planning.
• Provide a robust and maintainable tool for chemists.
• Facilitate the identification of purchasable precursors for target molecules.

Main Methods:

• Utilizes Monte Carlo tree search (MCTS) for recursive molecule decomposition.
• Employs an artificial neural network (ANN) policy to guide the search.
• Leverages a library of known chemical reaction templates.
• Developed with software engineering principles: testing, system design, CI.

Main Results:

• AiZynthFinder rapidly suggests precursors for retrosynthetic planning.
• Typically finds a solution in under 10 seconds.
• Completes a full search in under 1 minute.
• The software is robust, maintainable, and well-documented.

Conclusions:

• AiZynthFinder offers an efficient and user-friendly solution for retrosynthetic planning.
• The open-source nature and robust design make it suitable for both beginners and experienced researchers.
• Accelerates the process of identifying synthetic routes and starting materials.