Classification of organic reactions using similarity

Organic reaction classification allows for better understanding of organic reactivity, better data sharing, and better reaction storage and retrieval. The power of similarity has been often used to strengthen search methodology and to help reaction prediction systems, but seldom to optimise reaction classification. A hierarchical classification by similarity measures is proposed. It is based on electronic energy and chemical potential descriptors and on a general description of reactions. Using a main division into three reaction sets, additions, eliminations, and substitutions, followed by two successive subdivisions by number and types of reactive atoms it is possible to arrive at a hierarchy of groups containing homogeneous reaction classes. In addition, inside each set the reactions can be ordered. Some simple examples are used to demonstrate the efficiency of the classification and its overall consistency. The application of the final system to more complex examples is further proof of its efficiency. The reported classification can also be used in synthesis planning.