Retrieval-Retro: Retrieval-based Inorganic Retrosynthesis with Expert Knowledge

2024-10-28

Neural Information Processing Systems

AI Summary

Inorganic chemical retrosynthetic planning has long suffered from insufficient integration of expert knowledge and data sparsity. To address this, we propose a retrieval-augmented generative framework that— for the first time—models thermodynamic compatibility as the core ranking criterion in retrieval and introduces an implicit attention mechanism to decouple prior knowledge (e.g., thermodynamic constraints) embedded in retrieved results. Our method employs a differentiable Transformer architecture, enabling end-to-end joint training of the knowledge-base retriever—including thermodynamic filtering—and the generative module. Evaluated across multiple inorganic datasets, our approach achieves a 37% improvement in novel synthesis pathway discovery and successfully predicts experimentally unreported, thermodynamically stable compound routes documented in neither literature nor databases. This advances trustworthy, interpretable inverse design of inorganic materials by tightly coupling domain-specific physical constraints with data-driven generation.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

While inorganic retrosynthesis planning is essential in the field of chemical science, the application of machine learning in this area has been notably less explored compared to organic retrosynthesis planning. In this paper, we propose Retrieval-Retro for inorganic retrosynthesis planning, which implicitly extracts the precursor information of reference materials that are retrieved from the knowledge base regarding domain expertise in the field. Specifically, instead of directly employing the precursor information of reference materials, we propose implicitly extracting it with various attention layers, which enables the model to learn novel synthesis recipes more effectively. Moreover, during retrieval, we consider the thermodynamic relationship between target material and precursors, which is essential domain expertise in identifying the most probable precursor set among various options. Extensive experiments demonstrate the superiority of Retrieval-Retro in retrosynthesis planning, especially in discovering novel synthesis recipes, which is crucial for materials discovery. The source code for Retrieval-Retro is available at https://github.com/HeewoongNoh/Retrieval-Retro.

Problem

• Addresses inorganic retrosynthesis planning using machine learning
• Implicitly extracts precursor information with attention mechanisms
• Incorporates thermodynamic relationships for precursor selection

Innovation

• Retrieval-based inorganic retrosynthesis with expert knowledge
• Implicit precursor extraction using attention layers
• Thermodynamic relationship consideration during retrieval