Chemical Graph Mining for Classification of Chemical Reactions

The first step of prediction of product(s) of unknown chemical reactions by a computer is to classify known chemical reactions. Several authors came up with various reaction templates or reaction classes by studying (mostly) organic chemical reactions. This paper presents an algorithm that automates the process of finding reaction classes from known chemical reaction graphs. The algorithm presented here is sufficiently general to cover both organic and inorganic chemical reactions and can be applied to wider varieties of reactions. This graph mining algorithm derives all the reaction classes obtained from earlier schemes and thus earlier schemes are subsets of the present algorithm. The new reaction classes can also be derived from the algorithm presented in this paper.