Self-consistent assignment of asparagine and glutamine amide rotamers in protein crystal structures.

The current protein structure database contains unfavorable Asn/Gln amide rotamers in the order of 20%. Here, we derive a set of self-consistent potential functions to identify and correct unfavorable rotamers. Potentials of mean force for all heavy atoms are compiled from a database of high-resolution protein crystal structures. Starting from erroneous data, a refinement-correction cycle quickly converges to a self-consistent set of potentials. The refinement is entirely driven by the deposited structure data and does not involve any assumptions on molecular interactions or any artificial constraints. The refined potentials obtained in this way identify unfavorable rotamers with high confidence. Since the state of Asn/Gln rotamers is largely determined by hydrogen bond interactions, the features of the respective potentials are of interest in terms of molecular interactions, protein structure refinement, and prediction. The Asn/Gln rotamer assignment is available as a public web service intended to support protein structure refinement and modeling.