SynRoute: A Retrosynthetic Planning Software.

Area of Science:

• Computational chemistry
• Organic synthesis
• Artificial intelligence in chemistry

Background:

• Advancements in computer-assisted synthetic planning rely on large reaction databases and AI.
• Developing efficient tools for retrosynthetic analysis is crucial for drug discovery and chemical synthesis.

Purpose of the Study:

• To introduce SynRoute, a novel software tool for computer-assisted retrosynthetic planning.
• To develop a method for predicting the experimental feasibility of computer-generated chemical reactions using machine learning.

Main Methods:

• SynRoute utilizes 263 general reaction templates and a literature-based reaction database.
• Machine learning classifiers are trained on the Pistachio reaction database to predict reaction success.
• A vectorized Dijkstra-like search algorithm identifies and ranks synthetic routes by strategy.

Main Results:

• SynRoute successfully identified synthetic routes for an average of 83% of drug-like compounds from the ChEMBL database.
• Laboratory validation of 12 computer-generated routes confirmed their feasibility for synthesizing target compounds.

Conclusions:

• SynRoute provides a practical and efficient approach to computer-assisted synthetic planning.
• The integration of machine learning for reaction prediction enhances the reliability of computational retrosynthesis.