Improved statistics for determining the Patterson symmetry from unmerged diffraction intensities

Procedures for detecting the point-group symmetry of macromolecular data sets are examined and enhancements are proposed. To validate a point group, it is sufficient to compare pairs of Bragg reflections that are related by each of the group's component symmetry operators. Correlation is commonly expressed in the form of a single statistical quantity (such as R-merge) that incorporates information from all of the observed reflections. However, the usual practice of weighting all pairs of symmetry-related intensities equally can obscure the fact that the various symmetry operators of the point group contribute differing fractions of the total set. In some cases where particular symmetry elements are significantly under-represented, statistics calculated globally over all observations do not permit conclusions about the point group and Patterson symmetry. The problem can be avoided by repartitioning the data in a way that explicitly takes note of individual operators. The new analysis methods, incorporated into the program LABELIT (http://cci.lbl.gov/labelit), can be performed early enough during data acquisition, and are quick enough that it is feasible to pause to optimize the data collection strategy.