Abstract
In silico protein-ligand docking methods have proven to be useful in drug design and have also shown promise for predicting the substrates of enzymes, an important goal given the number of enzymes with uncertain function. Further testing of this latter approach is critical because (1) metabolites are on average much more polar than druglike compounds and (2) binding is necessary but not sufficient for catalysis. Here, we demonstrate that docking against the enzymes that participate in the 10 major steps of the glycolysis pathway in Escherichia coli succeeds in identifying the substrates among the top 1% of a virtual metabolite library.