Table 1. Correlation of density-modified phases with true phases [〈cos(Δϕ)〉] for model data in a unit cell containing 32–68% solvent.

Data and analysis using reciprocal-space solvent flattening are from Terwilliger (1999 ▶). Phases with simulated errors for 6906 data from ∞ to 3.0 Å were constructed using a model consisting of coordinates from a dehalogenase enzyme from Rhodococcus species ATCC 55388 (American Type Culture Collection, 1992 ▶) determined recently in our laboratory (Newman et al., 1999 ▶; PDB entry 1bn7), except that some of the atoms were not included in order to vary the fraction of solvent in the unit cell. The cell was in space group P2₁2₁2, with unit-cell parameters a = 94, b = 80, c = 43 Å and one molecule in the asymmetric unit. Phases with simulated errors were generated by adding phase errors as described in Terwilliger (1999 ▶) to yield an average value of the cosine of the phase error (i.e. the true figure of merit of the phasing) of 〈cos(Δϕ)〉 = 0.42 for acentric and 0.39 for centric reflections. The model data with simulated errors was then density modified by the maximum-likelihood method described here, by reciprocal-space solvent flattening (Terwilliger, 1999 ▶) and by a real-space method as implemented in the program dm (Cowtan & Main, 1996 ▶), version 1.8, using solvent flattening and histogram matching.

Fraction protein (%)	Starting	Real-space solvent flattening	Reciprocal-space solvent flattening	Maximum-likelihood solvent flattening
32	0.41	0.64	0.85	0.87
42	0.40	0.62	0.67	0.83
50	0.41	0.54	0.56	0.77
68	0.42	0.48	0.41	0.53