Solvent model for protein crystals: on occupancy parameters for discrete solvent sites and the solvent continuum

On the basis of test calculations, Kundrot & Richards [Acta Cryst. (1987), B43, 544-547] suggest that for models of protein crystal structures based on limited-resolution X-ray data, it is not appropriate to vary both the occupany (Q) and thermal (B) parameters for the solvent molecules during refinement. There are compelling reasons, however, to adjust both parameters. If the data are sufficiently extensive to include a B parameter for each solvent site, then an adjustable Q parameter should also be included if the model is to represent physical reality. To fix Q at some arbitrary value as suggested, means that differences in occupancy will be absorbed in the 'thermal' parameters, leading to the errors corresponding to the electron density plots of Kundrot & Richards. Although the errors appear to be relatively small, they will be accentuated in Fourier maps of complex structures based on real data, causing error peaks, both positive and negative, that may greatly exceed the random errors. The current practice of using the scattering factor of the O atom, fO, for the water molecule neglects the scattering from the H atoms. A modified scattering factor based on fO- would approximate the scattering from water molecules better. Finally, the solvent continuum should be included in the model, and the low-order data included in calculating the Fourier maps from which the Q parameters are estimated.