Metabolic pathway databases such as KEGG contain information on thousands of enzymatic reactions drawn from the biomedical literature.  Ensuring consistency of such large metabolic pathways is essential to their proper use and, in particular, the problem of determining the optimal atom to atom mapping between the substrate and the product of an enzymatic reaction, is fundamental to ensuring consistency of the enzymatic reactions that are stored in a metabolic pathway database.  In our approach, the knowledge of the atomic rearrangement pattern in enzymatic reactions is exploited in order to reduce the automatic atom mapping problem to a series of chemical substructure searches between the substrate and the product of the enzymatic reaction.  Our approach will be illustrated in this talk by an exhaustive validation of a substantial portion of the KEGG LIGAND database.