Abstract
It is fundamental for science and technology to be able to predict chemicalreactions and their properties. To achieve such skills, it is important todevelop good representations of chemical reactions, or good deep learningarchitectures that can learn such representations automatically from the data.There is currently no universal and widely adopted method for robustlyrepresenting chemical reactions. Most existing methods suffer from one or moredrawbacks, such as: (1) lacking universality; (2) lacking robustness; (3)lacking interpretability; or (4) requiring excessive manual pre-processing.Here we exploit graph-based representations of molecular structures to developand test a hypergraph attention neural network approach to solve at once thereaction representation and property-prediction problems, alleviating theaforementioned drawbacks. We evaluate this hypergraph representation in threeexperiments using three independent data sets of chemical reactions. In allexperiments, the hypergraph-based approach matches or outperforms otherrepresentations and their corresponding models of chemical reactions whileyielding interpretable multi-level representations.