Aminoacyl-tRNA synthetases guard the fidelity of cognate amino acid incorporation during protein biosynthesis; for example, phenylalanyl-tRNA synthetase (PheRS) activates and transfers only Phe to its tRNA. Since we are interested in using a computational protocol to identify nonnatural amino acids that are incorporated by wild-type PheRS, it is critical to understand the fidelity of PheRS in binding the 20 natural amino acids. To this end, Hier-Dock, a computational protocol for predicting binding sites and relative binding affinities, was used for testing the natural amino acids in PheRS. Scanning the entire ligand-accessible protein surface for the best binding region, we find that HierDock correctly identifies the active site of Phe in PheRS and predicts Phe within 0.61 Angstrom RMSD of the crystal structure. HierDock also successfully shows PheRS discriminates for Phe, as the noncognate amino acids bind less favorably in the binding site of Phe. However, we find that Met, Cys, and Tyr bind competitively but at positions distant from the Phe binding site. This result corroborates in vitro measurements of aminoacyl adenylate formation, which show Met competes with Phe at the amino acid binding stage. We predict that the binding site of Met would not activate PheRS, as the noncognate amino acid cannot establish suitable hydrogen bonds with the PheRS reaction center. These results validate the use of HierDock in predicting the binding sites of the cognate amino acids in PheRS. The HierDock procedure calculates the discrimination of aminoacyl-tRNA synthetases at the stage of binding the cognate amino acid and offers a molecular level understanding of the mistakes made in protein biosynthesis that are not readily uncovered through experiments. This technique is also useful for predicting the binding of a selected nonnatural amino acid analogue, thereby indicating whether the molecule would be incorporated into a wild-type aminoacyl-tRNA synthetase.