Metabolic Pathway Modeling by Using the Nearest Neighbor Algorithm

A new computational approach was developed for modeling the metabolic pathways. The new approach is featured by combing the knowledge of gene ontology, microarray, and chemical functional group to formulate the enzyme-substrate/product couples in a 1,660 vector space. The nearest neighbor algorithm was used to perform the prediction of the networking relationship occurring in the metabolic pathways. The average overall success rate by jackknife cross-validation tests for the 79 metabolic pathways in the budding yeast system was over 94%, suggesting that the current approach might become a useful tool for studying metabolic pathways and many other networking-related areas.