In protein crystallography, much time and effort are often required to trac
e an initial model from an interpretable electron density map and to refine
it until it best agrees with the crystallographic data. Here, we present a
method to build and refine a protein model automatically and without user
intervention, starting from diffraction data extending to resolution higher
than 2.3 Angstrom and reasonable estimates of crystallographic phases. The
method is based on an iterative procedure that describes the electron dens
ity map as a set of unconnected atoms and then searches for protein-like pa
tterns. Automatic pattern recognition (model building) combined with refine
ment, allows a structural model to be obtained reliably within a few CPU ho
urs. We demonstrate the power of the method with examples of a few recently
solved structures.