RAscore

RAscore predicts retrosynthetic accessibility of small molecules by classifying whether AiZynthFinder can identify a synthetic route.

Key Features:

• Binary Classification: RAscore outputs a binary score (1 or 0) indicating whether AiZynthFinder can find a synthetic route for a target molecule.
• Machine Learning Models: Models are trained to replicate AiZynthFinder predictions and produce retrosynthetic accessibility assessments (RAscores).
• Speed and Efficiency: RAscore computes retrosynthetic accessibility scores approximately 4,500 times faster than running retrosynthetic searches directly with AiZynthFinder.
• Integration with Virtual Screening Workflows: RAscore can pre-screen millions of virtual molecules from enumerated or generative models to prioritize compounds with feasible synthetic routes for downstream virtual screening.

Scientific Applications:

• Drug Discovery: Rapidly assess synthesizability to prioritize candidate molecules during early-stage drug discovery and optimize resource allocation.
• Chemical Biology and Library Design: Filter and curate large virtual libraries to build databases enriched for compounds with practical synthetic routes for biological screening.

Methodology:

Machine learning models trained on AiZynthFinder predictions classify whether a synthetic route can be identified and output a binary retrosynthetic accessibility score (RAscore).