No abstract

A two-channel maximum-entropy method (MEM), first used to enhance magnetization densities from phased polarized neutron data by Papoular & Gillon [(1990). Europhys. Lett. 13,[429][430][431][432][433][434], has been applied to the electron deformation density. The resulting entropic densities are compared with standard deformation densities and with dynamic and static deformation maps obtained from multipole ref'mements. The procedure is illustrated with simulated and real single-crystal X-ray data sets on the molecular crystal of ot-glycine. Both a uniform prior and a prior equal to the MEMenhanced dynamic model deformation density are used in the MEM procedure, the result of which does not depend on the starting density. The method is judged by the appearance of the resulting maps and the values of the molecular dipole moment before and after the MEM. Compared with the conventional deformation density, the MEM procedure sharpens the peaks in the bond but flattens the weaker features, especially when a uniform prior is used. The dipole-moment criterion shows the non-uniform prior to be preferable to the uniform prior in reproducing electrostatic properties. The usefulness of the MEM in charge-density analysis remains open to discussion.

Very strong reflections have a dominant impact on the initial phasing and model-building stages of structure determination. However, experimental phasing (MIR, SAD or MAD) fails on some of the strongest reflections when the heavy-atom contribution to scattering is relatively weak or absent. It is shown that when just a few (approximately 50-100) of these reflections are assigned low-error phases, the entire set of isomorphous replacement phases becomes significantly improved after density modification. This improvement is indicated by higher map correlation coefficients and reduced mean phase errors of the updated data. The problem of phasing the strongest reflections may be solved by the direct measurement of triplet phases in a three-beam diffraction experiment. The analysis shows that merging isomorphous replacement data with a limited number of highly accurate phases from the reference-beam diffraction experiment would significantly improve conventional experimental phasing.

Conventional experimental phasing methods are most accurate for moderate-resolution reflections, with progressively greater ambiguity in the phases of reflections away from this optimal point. Frequently, very strong (usually low-resolution) reflections are either poorly phased or altogether unrecorded. While the spatial frequency of these reflections is predominantly too low to dramatically affect the calculated electron density at an atomic level, they have a dominant impact on the determination of the large-scale distribution of matter in the unit cell. Consequently, while these few strong reflections play only a peripheral role in the latter stages of a structure-determination project, they are crucial to the success of initial phasing and model-building efforts. Here, the pivotal importance of a limited number of strong/low-resolution reflection phases is shown and a procedure to derive these phases is described. The improvement in map correlation coefficients after density modification of a marginal ;starting' MAD data set (obtained from two Zn atoms at special positions in rhombohedral insulin crystals) was compared with the improvement in map correlation coefficients observed after density modification of an ;expanded' data set obtained by combining a limited number of highly accurate phases measured using three-beam diffraction with the ;starting' MAD data. It is concluded that a small number of high-amplitude/low-resolution reflections contribute disproportionately to generating an initial structure and it is suggested that a small number of triplet phases could be measured quickly and combined with experimental isomorphous replacement phases in order to move stubborn structures for novel proteins down the structure-solution pathway.

A new phase-extension procedure has been applied to isomorphous replacement data and shown to yield improved phases and maps compared with standard solvent¯attening operating on a full set of centroid phases. In this procedure, a starting subset of core phases is selected based on the sharpness of the phase-probability curves. Phase extension using solvent¯attening as the density-modi®cation procedure is then carried out, gradually adding additional phases. In tests with known protein structures, the mean phase errors for the output expanded phase sets were reduced by 3±9 and the corresponding map correlation coef®cients were increased by 0.05±0.18 relative to phase sets from standard solvent-¯attening procedures. With SIR data, the lowest ®nal mean phase errors were approximately 58 and the corresponding map correlation coef®cients were in the range 0.53±0.68.