How good are my data and what is the resolution?

Abstract

Following integration of the observed diffraction spots, the process of `data reduction' initially aims to determine the point-group symmetry of the data and the likely space group. This can be performed with the program POINTLESS. The scaling program then puts all the measurements on a common scale, averages measurements of symmetry-related reflections (using the symmetry determined previously) and produces many statistics that provide the first important measures of data quality. A new scaling program, AIMLESS, implements scaling models similar to those in SCALA but adds some additional analyses. From the analyses, a number of decisions can be made about the quality of the data and whether some measurements should be discarded. The effective `resolution' of a data set is a difficult and possibly contentious question (particularly with referees of papers) and this is discussed in the light of tests comparing the data-processing statistics with trials of refinement against observed and simulated data, and automated model-building and comparison of maps calculated with different resolution limits. These trials show that adding weak high-resolution data beyond the commonly used limits may make some improvement and does no harm.

Keywords: data reduction; data scaling; data statistics; software.

Figure 1

New graphs from AIMLESS against ‘batch’ or image number. (a, c) Nominal resolution estimated as the point at which 〈I/σ(I)〉 falls below 1.0, showing a trend to lower resolution with increasing radiation damage, with both values for individual batches and values smoothed over a 5° range. (b, d) Cumulative completeness for all data and anomalous differences. (a) and (b) show that in this good case the damaged data in the last third of the sweep can be safely discarded without reducing the completeness. (c) and (d) show graphs for a poor and incomplete data set from two crystals. At the end of this data collection the anomalous data are still very incomplete. Breaks in the x axis separate the two crystals.

Figure 2

Plots of data-processing and refinement statistics against resolution. In (a), (b), (e) and (f), CC_1/2 is shown as a dashed red line and 〈〈I〉/σ(〈I〉)〉 is shown as a pale blue dotted line (right-hand axes). (a), (b), (c) and (d) are from example 1, (e) is from example 4 and (f) is from example 5. (a) Data-processing statistics and R _free for the observed data (black); against simulated data beyond 2.4 Å resolution, expected |F| (blue); and |F| from two or three sets of random intensities (green). (b) Similar statistics in cones around the three principal directions of anisotropy d1 (red), d2 (= b*, black) and d3 (blue), omitting the F(random) values. (c) R _free(F from I) values for refinement against measured and random simulated intensities. (d) ML scale that indicates the contribution of these reflections to electron-density calculations. (e) As (a) but for example 4. (f) The same for example 5

Figure 3

OMIT difference maps (mF _o − DF _c) at different resolutions, along with data-processing statistics plotted against resolution. (a) A sugar (NAG) chain from example 4. (b) An omitted residue from example 1.