DIALS as a toolkit

Abstract

The DIALS software for the processing of X-ray diffraction data is presented, with an emphasis on how the suite may be used as a toolkit for data processing. The description starts with an overview of the history and intent of the toolkit, usage as an automated system, command-line use, and ultimately how new tools can be written using the API to perform bespoke analysis. Consideration is also made to the application of DIALS to techniques outside of macromolecular X-ray crystallography.

Keywords: X-ray crystallography; methods development; open source; software.

FIGURE 1

Schematic rocking curve of a single reflection, illustrating the undercounting and count rate correction in a pixel array detector. Horizontal levels indicate the mean true incident flux in each exposure (red), the mean observed flux in each exposure (blue), and the effect on the observed flux of the detector's in‐built count rate correction (black)

FIGURE 2

Simple workflow of data processing, which is largely independent of the processing package used

FIGURE 3

Simple workflow of data processing with DIALS, which deliberately follows the simple workflow shown in Figure 2

FIGURE 4

Zoomed‐in views of four modules of an EIGER 2XE: (a) a single image, (b) the maximum pixels from a 10 image “stack” (corresponding to an image width of 1°), (c) the same with the “shoeboxes” from spot finding superimposed and (d) the maximum value of pixels in a 25 image stack

FIGURE 5

DIALS image viewer (top) and reciprocal lattice viewer (bottom), showing the outcome of spot finding on a data set from a crystal of cubic insulin, recorded on a DECTRIS EIGER 2XE 9M detector. The image viewer has 10 images stacked, giving 2° of rotation, while the reciprocal lattice view is the entire data set

FIGURE 6

Scan‐varying cell parameters (top) and orientation (bottom) of cubic insulin rotated through 3,600°. The best‐fit unit cell parameters clearly vary depending on the observed angle. This could be due to variations in crystal uniformity as it rotates through the beam

FIGURE 7

CC½ (top) and R _meas (bottom) for both profile‐fitted and summation integrated cubic insulin data. The data from profile fitting have both a superior half set correlation and lower merging residual at high resolution, where the reflections are at their weakest. At low resolution the figures are similar for both methods

FIGURE 8

Fraction of reflections indexed in a macromolecular case with a strong single lattice (top), and a small molecule data set with two distinct lattices (bottom). In the upper image, it is clear that the vast majority of reflections are indexed. In the lower, the fact that about half are indexed suggests that there is a second (unindexed) lattice in the data set

FIGURE 9

View captured from the image viewer showing a zoomed‐in view of reflection shoeboxes from integrating a two‐lattice small molecule data set. The boxes correspond to the entire reflection, including background with the red and blue boxes represent different lattices. The detail of which pixels are peak and background are not saved by default

FIGURE 10

Reciprocal lattice view of a data set with two lattices present, showing the relative orientation of the reciprocal cells

FIGURE 11

Pixel intensities drawn from strong pixels from two data sets collected in an apparently equivalent manner, though the faster collection clearly saturates the detector