A method for refinement of partially interpreted protein structures including a procedure for scaling between a model and an electron-density map

A method for the structure-factor least-squares refinement of a poor and incomplete atomic model of a protein molecule is suggested. Usually structure-factor least squares treat calculated phases as an observable to be associated with observed amplitudes of the complete model and assume that the difference between observed and calculated structure factors arises from the inaccuracies of the model. It is proposed that the assumption may be made valid for an incomplete model by including, in the calculation of the structure factors, electron-density features observed in the region outside the existing model. Structure factors obtained from the atomic model and from electron density may be scaled by fitting the radial distribution function in reciprocal space and their relative weights may be determined by an R-factor search procedure.