AiZynthFinder

AiZynthFinder performs retrosynthetic planning to deconstruct target molecules into simpler, purchasable precursors using a Monte Carlo tree search guided by an artificial neural network policy and reaction templates.

Key Features:

• Monte Carlo Tree Search (MCTS): Uses MCTS to recursively break down molecular structures into candidate disconnections.
• Artificial neural network policy: Employs a neural network policy to suggest potential precursor molecules for each disconnection.
• Reaction template library: References a library of known reaction templates to propose template-based transformations.
• Performance: Typically identifies viable synthetic routes in under 10 seconds and can complete exhaustive searches within one minute.
• Object-oriented architecture: Implements an object-oriented design that enables extension and incorporation of new features.
• Software engineering practices: Incorporates automatic testing, systematic design, and continuous integration to ensure reliability.

Scientific Applications:

• Retrosynthetic route generation: Deconstructs target molecules into sequences of reactions leading to purchasable precursors.
• Rapid synthesis planning: Provides fast identification of candidate synthetic routes for experimental planning and evaluation.
• Method development and extension: Serves as a platform for integrating new retrosynthetic methodologies and reaction data.

Methodology:

Performs a recursive Monte Carlo tree search (MCTS) guided by an artificial neural network policy that suggests precursors by referencing a library of known reaction templates.