Assessing Data Quality

Why

The first step after data processing is analysis of the diffraction data to assess quality and detect any pathologies that might make structure solution difficult. Data quality can be quantified in terms of resolution of diffraction, anomalous signal-to-noise, and consistency with prior knowledge about diffraction from crystals.

Possible pathologies include twinning, translational non-crystallographic symmetry, anisotropy, and missed symmetry elements. For example, crystals are considered ‘twinned’ if two or more separate crystals (domains) are intergrown in such a way that they share some crystal lattice points in a symmetrical manner. And translational noncrystallographic symmetry (tNCS) occurs when two or more independent copies of a molecule or assembly have a similar orientation in the asymmetric unit of the crystal.

How

The phenix.xtriage program analyzes data to assess quality and detect possible problems. For most uses, the program only requires a reflection file containing the data set you wish to analyze. Running phenix.xtriage generates information about the data set, which is best viewed in the GUI as graphs and tables. The first section of the output provides an overall assessment of the data.

How to use the phenix.xtriage GUI: Click here

Phenix reference manual for phenix.xtriage

Common issues

Interpretation of twinning results: There are two parts to testing for twinning — determining whether the overall intensity statistics look normal, and testing for specific twinning operations. Both kinds of tests need to indicate twinning for the data set to be considered twinned. In some cases, the tests of specific twinning operations return a high twin fraction but the overall intensity statistics look normal; this should not be interpreted as twinning. More likely, a symmetry element is missing, and phenix.xtriage might suggest a higher symmetry.

Related programs and Documentation

• phenix.plan_sad_experiment: This tool allows you to plan your SAD data collection strategy by estimating the anomalous signal and the probability of solving your structure with SAD based on the anomalously-scattering atoms, wavelength, size of your molecule, and overall I/sigI for your dataset.
• phenix.anomalous_signal: This tool is like phenix.plan_sad_experiment except that it uses your measured data to give much more accurate estimates. Specifically, it uses your scaled data, two half-dataset files of your scaled data, the name of the anomalously-scattering atom, and the number of sites in the substructure in order to estimate the anomalous signal in your data and the probability of solving your structure by SAD. You can create the necessary scaled half-datasets using phenix.scale_and_merge.
• HKLviewer: This program visualizes reciprocal space data. This can be very useful for detecting systematic problems, such as incompleteness (e.g., missing cones or wedges), TNCS or twinning.
• phenix.explore_metric_symmetry: This program tests whether your crystal lattice is consistent with other lattices/symmetries. It also can test whether two lattices are related. (Command line only).
• Using unmerged data in Phenix: You can analyze unmerged data in Phenix to provide useful information, including the new CC1/2 and CC* statistics. The program phenix.merging_statistics calculates these and a number of other statistics. The program phenix.scale_and_merge scales unmerged data for you.