🤖 AI Summary
This work addresses the challenge of reaction yield prediction, which is hindered by scarce labeled data, the vast and sparse reaction space, and the inability of existing representations to capture complex chemical transformations. To overcome these limitations, the authors propose RxnCLF, a self-supervised contrastive learning framework that introduces a novel condensed reaction graph (CRG) integrating both reactant and product information. By leveraging graph neural networks, RxnCLF learns explicit and interpretable transformation structures and models chemical reactions within a unified continuous latent space. The method significantly outperforms current graph- and sequence-based models on multiple yield prediction benchmarks, demonstrating substantial improvements in R² scores, and exhibits strong generalization capabilities on downstream tasks such as regioselectivity and enantioselectivity prediction, offering a new paradigm toward a general-purpose foundation model for chemical reactions.
📝 Abstract
Reaction yield prediction remains challenging because labeled data are scarce and reaction space is both combinatorially large and sparsely populated, limiting the generalization of existing reaction representations. String-, fingerprint-, and graph-based reaction encodings only partially capture chemical transformations, making accurate prediction difficult for reactions with complex substrates. We propose reaction contrastive learning foundation (RxnCLF), a self-supervised contrastive framework for reaction representation learning. RxnCLF is built on a condensed reaction graph (CRG) that unifies reactant and product information into a single graph, enabling the model to learn explicit and enriched transformation structure rather than disconnected graphs. Pretrained on 1.7 million Pistachio reactions, RxnCLF learns a compact and continuous latent space that captures both reaction-center features and broader side chain contexts, making it transformation-aware and chemically interpretable. Fine-tuned on multiple yield prediction benchmarks, including Buchwald-Hartwig, Pd-catalyzed BH coupling, and proprietary HTE C-N coupling and amide formation datasets, RxnCLF consistently outperforms graph and sequence-based baselines, improving R2 and achieving the best performance overall. Our results highlight the promise of CRG-based RxnCLF as a scalable reaction foundation model, with the potential to generalize across broader reaction spaces and support diverse downstream reaction informatics tasks, including regioselectivity prediction, enantioselectivity prediction, and reaction condition optimization.