ASKCOS: Open-Source, Data-Driven Synthesis Planning

Area of Science:

• * Computational chemistry and cheminformatics.
• * Organic synthesis and reaction informatics.

Background:

• * Machine learning and large reaction datasets have accelerated data-driven computer-aided synthesis planning (CASP).
• * Bridging the gap between research and practical application is crucial for advancing chemical synthesis.

Purpose of the Study:

• * To describe data-driven methods and models in the newest version of ASKCOS, an open-source synthesis planning software.
• * To highlight the integration of various modules for comprehensive synthesis planning capabilities.

Main Methods:

• * Integration of modules for retrosynthetic planning (Interactive Path Planner, Tree Builder with MCTS and Retro*).
• * Utilization of four one-step retrosynthesis models (template-based and template-free).
• * Development of capabilities for reaction condition, pathway, and outcome prediction (major product, impurities, selectivity).

Main Results:

• * ASKCOS integrates retrosynthesis, condition prediction, and outcome prediction modules.
• * Auxiliary capabilities include solubility and quantum mechanical descriptor prediction.
• * The software provides user-friendly interfaces for chemists.

Conclusions:

• * ASKCOS assists chemists in daily tasks by complementing expert decision-making and route ideation.
• * Computer-aided synthesis planning tools are increasingly important and accessible in modern chemistry research.
• * The integration of diverse data-driven methods enhances the utility of CASP software.