Comparison of Bayesian and regression models in missing enzyme identification

Computational identification of missing enzymes is important in metabolic network reconstruction. For a metabolic reaction, given a set of candidate enzymes identified by biological evidences, a powerful predictive model is necessary to predict the actual enzyme(s) catalysing the reaction. In this study, we compare Bayesian Method, which is used in previous work, with several regression models. We apply the models to known reactions in E. coli and three other bacteria. It is shown that the proposed regression models obtain favourable performance when compared with the Bayesian method.