Improving route development using convergent retrosynthesis planning

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Organic Synthesis

Background:

• Computer-aided synthesis planning typically generates individual routes from target molecules to building blocks.
• Existing methods do not efficiently leverage shared synthetic pathways for multiple targets, limiting library synthesis in medicinal chemistry.
• Convergent synthesis, which maximizes shared intermediates, is crucial for efficient library production.

Purpose of the Study:

• To develop a novel computational approach for planning convergent retrosynthetic routes for multiple target molecules simultaneously.
• To enhance the efficiency and practical applicability of retrosynthetic planning in medicinal chemistry library synthesis.
• To identify and analyze complex synthesis routes with shared intermediates from real-world datasets.

Main Methods:

• Utilized a graph-based processing pipeline to analyze Johnson & Johnson Electronic Laboratory Notebooks (J&J ELN) and public datasets.
• Developed a multi-target, multi-intermediate planning approach guided by machine learning single-step retrosynthesis models.
• Evaluated the approach by identifying convergent routes and assessing their solvability.

Main Results:

• Over 70% of reactions analyzed were part of convergent synthesis, covering over 80% of J&J ELN projects.
• The novel planning approach identified convergent routes for over 80% of test cases with individual compound solvability exceeding 90%.
• Convergent search enabled synthesizing nearly 30% more compounds simultaneously compared to individual search methods, increasing common intermediate utilization.

Conclusions:

• The developed convergent synthesis planning approach significantly enhances efficiency and practical applicability in medicinal chemistry.
• Convergent retrosynthetic strategies are prevalent and highly effective, with the new method demonstrating high solvability and increased compound yield.
• This approach offers a powerful tool for optimizing library synthesis by maximizing convergence and shared intermediates.