Abstract:Motivation: Retrosynthesis planning poses a formidable challenge in the organic chemical industry. Single-step retrosynthesis prediction, a crucial step in the planning process, has witnessed a surge in interest in recent years due to advancements in AI for science. Various deep learning-based methods have been proposed for this task in recent years, incorporating diverse levels of additional chemical knowledge dependency.

What problem does this paper attempt to address?

The paper primarily proposes a new solution to the problem of retrosynthesis in organic chemistry. Retrosynthesis is a key task in the organic chemical industry that involves finding an efficient synthetic route from a target molecule to accessible starting materials, which is particularly challenging in the pharmaceutical industry. To address this issue, computer-aided synthesis planning (CASP) has garnered increasing attention in recent years. The method introduced in the paper is called UAlign, a template-free graph-to-sequence pipeline for single-step retrosynthesis prediction. The main contribution of this method lies in overcoming the limitations of Transformer-based methods in molecular representation learning and the insufficient utilization of chemical information. Specifically, UAlign combines the advantages of Graph Neural Networks (GNN) and Transformer models, enabling more effective use of the inherent graph structure information of molecules. Additionally, based on the fact that most molecular structures remain unchanged during chemical reactions, the paper proposes a simple yet effective SMILES alignment technique to facilitate the reuse of unchanged structures, thereby generating reactants. A key innovation of UAlign is the introduction of an unsupervised learning mechanism to establish the correspondence between product and reactant SMILES tokens, which has achieved even better results than supervised SMILES alignment methods. Through extensive experimental validation, UAlign significantly outperforms state-of-the-art template-free methods and, in some cases, can rival or surpass template-based methods. In summary, UAlign aims to address the limitations of current retrosynthesis prediction methods, particularly for template-free methods. It enhances model performance and can substantially leverage the structural information in product molecules to guide the generation of reactants.

UAlign: Pushing the Limit of Template-free Retrosynthesis Prediction with Unsupervised SMILES Alignment

Retrosynthesis prediction with an interpretable deep-learning framework based on molecular assembly tasks

RetroComposer: Composing Templates for Template-Based Retrosynthesis Prediction

Improve retrosynthesis planning with a molecular editing language

RetroPrime: A Diverse, Plausible and Transformer-based Method for Single-Step Retrosynthesis Predictions

RetroPrime: A Chemistry-Inspired and Transformer-based Method for Retrosynthesis Predictions

Bridging Chemical Knowledge and Machine Learning for Performance Prediction of Organic Synthesis.

RetroXpert: Decompose Retrosynthesis Prediction like a Chemist

Automatic Retrosynthetic Pathway Planning Using Template-free Models

Retrosynthesis prediction using an end-to-end graph generative architecture for molecular graph editing

Retrosynthesis Prediction with Local Template Retrieval

Site-Specific Template Generative Approach for Retrosynthetic Planning

Retrosynthesis Prediction with an Iterative String Editing Model

Root-aligned SMILES: a Tight Representation for Chemical Reaction Prediction

Application of Modern Intelligent Algorithms in Retrosynthesis Prediction

Retroformer: Pushing the Limits of Interpretable End-to-end Retrosynthesis Transformer

Artificial Intelligence for Retrosynthesis Prediction

Recent advances in deep learning for retrosynthesis

Deep Retrosynthetic Reaction Prediction using Local Reactivity and Global Attention

Node-Aligned Graph-to-Graph (NAG2G): Elevating Template-Free Deep Learning Approaches in Single-Step Retrosynthesis

Challenging Complexity with Simplicity: Rethinking the Role of Single-Step Models in Computer-Aided Synthesis Planning