Reliable quality-control methods for protein crystal structures

The emergence of structure‐determination initiatives that employ high‐throughput protein crystallography emphasizes the need to establish quality‐control methods for screening the resulting models prior to deposition with the public data banks. An in‐house database of 26 new protein structures, associated diffraction data and high‐quality experimentally determined electron‐density maps have been used to develop (i) a set of minimal global quality criteria that a structure must meet before the refinement may be considered completed and (ii) a reliable set of indicators for detecting local errors in protein structures. These criteria have been applied to detecting local errors to a set of structures recently deposited in the Protein Data Bank and it is estimated that about 3% of amino acids are incorrectly modeled.