Reaction Classification Study Using Reaction Similarity (ReacSim) Vector for Recognition of Similarities in Chemical Reactions

Abstract

A new representation method of chemical reactions ReacSim (Reaction Similarity) vector is described. The ReacSim vector has been developed for systematic classification of chemical reaction, about which we are studying for organizing chemical reactions and systematic derivation of knowledge of reactions. The ReacSim vector is defined as highest deviation of physico-chemical parameters from reactants and products. The ReacSim vectors are independent of reaction sites, namely, independent on the number of atoms and bonds of the reaction sites. This benefit of the ReacSim vector makes it possible to classify variety of chemical reactions, generally. 131 reactions that are represented by the ReacSim vectors are classified by a self-organizing neural network method Kohonen neural network. Good correlation between the ReacSim vectors and types of reactions is shown by the classification studies.