Validation of protein models from Calpha coordinates alone.

The Protein Data Bank contains a number of structures for which only the coordinates of the Calpha atoms have been deposited. Although many tools are available for the validation of all-atom protein models, hardly any of these can be used to assess the quality of models for which only Calpha coordinates are available. Two rapid and simple tests to assess the quality of the Calpha backbone of a protein model are described, one based on the distribution of Calpha-Calpha distances, and the other on the two-dimensional distribution of the angles and dihedrals formed by sequential Calpha atoms. Expected distributions were derived by analysing a set of 1343 high-resolution, all-atom protein models. The distance criterion is useful to discriminate between refined and unrefined models, whereas the angle/dihedral criterion can be used to discriminate between normal and possibly problematic Calpha models. The method has been applied to a set of 88 Calpha-only models from the Protein Data Bank. The tracing of two of the models that are outliers in this analysis has recently been shown to be incorrect. Other applications of the method are discussed.