MechRetro is a chemical-mechanism-driven graph learning framework for interpretable retrosynthesis prediction and pathway planning

Abstract

Leveraging artificial intelligence for automatic retrosynthesis speeds up organic pathway planning in digital laboratories. However, existing deep learning approaches are unexplainable, like "black box" with few insights, notably limiting their applications in real retrosynthesis scenarios. Here, we propose MechRetro, a chemical-mechanism-driven graph learning framework for interpretable retrosynthetic prediction and pathway planning, which learns several retrosynthetic actions to simulate a reverse reaction via elaborate self-adaptive joint learning. By integrating chemical knowledge as prior information, we design a novel Graph Transformer architecture to adaptively learn discriminative and chemically meaningful molecule representations, highlighting the strong capacity in molecule feature representation learning. We demonstrate that MechRetro outperforms the state-of-the-art approaches for retrosynthetic prediction with a large margin on large-scale benchmark datasets. Extending MechRetro to the multi-step retrosynthesis analysis, we identify efficient synthetic routes via an interpretable reasoning mechanism, leading to a better understanding in the realm of knowledgeable synthetic chemists. We also showcase that MechRetro discovers a novel pathway for protokylol, along with energy scores for uncertainty assessment, broadening the applicability for practical scenarios. Overall, we expect MechRetro to provide meaningful insights for high-throughput automated organic synthesis in drug discovery.

Accelerating the integration of ChatGPT and other large‐scale AI models into biomedical research and healthcare

“… DrugEx v3 : A drug design approach that utilizes scaffold constraints and reinforcement learning based on graph transformers 94 MechRetro : A graph learning framework that employs chemical mechanisms to predict and plan pathways for retrosynthesis in an interpretable manner 95 MHTAN‐DTI : A hierarchical transformer and attention network that uses metapaths to predict interactions between drugs and targets 96 …”

“…MechRetro : A graph learning framework that employs chemical mechanisms to predict and plan pathways for retrosynthesis in an interpretable manner 95 …”

Advanced deep learning methods for molecular property prediction

“…Recently, artificial intelligence (AI) technologies, represented by deep learning, have achieved great success in biology and chemistry fields [3][4][5], which indicates the great potential and outstanding creativity of AI. Compared to traditional drug property prediction methods, deep learning models can eliminate the complicated feature engineering process, accelerate computation process, and increase availability of large data sets.…”