Data augmentation in a triple transformer loop retrosynthesis model

Key Points

• The research aims to enhance computer-assisted synthesis planning by mitigating biases in reaction types using a data augmentation approach.
• Developed a triple transformer loop model for retrosynthesis
• Implemented data augmentation techniques to diversify training data
• Analyzed reaction types from the US Patent Office database
• Demonstrated improved model performance in predicting diverse reaction types
• Reduced bias in reaction selection during synthesis planning
• Enhanced capability for computer-assisted synthesis from varied data sources