Optimization of data collection taking radiation damage into account

Abstract

To take into account the effects of radiation damage, new algorithms for the optimization of data-collection strategies have been implemented in the software package BEST. The intensity variation related to radiation damage is approximated by log-linear functions of resolution and cumulative X-ray dose. Based on an accurate prediction of the basic characteristics of data yet to be collected, BEST establishes objective relationships between the accessible data completeness, resolution and signal-to-noise statistics that can be achieved in an experiment and designs an optimal plan for data collection.

Figure 1

(a) Wilson plots (average observed J in resolution shells) for P19–siRNA data. Black and grey squares correspond to the fresh crystal and to the crystal after an absorbed dose of 30 MGy, respectively. The BEST-predicted Wilson plots, i.e. the average values of calculated for the same set of reflections, are represented by solid and dashed lines, respectively. (b) Relative scale and B factors as a function of radiation dose. Isotropic B-factor scaling to a common reference scale is performed by BEST. The scale factors are divided by those of the first data set and the B factor of the first data set is subtracted from the B factors of subsequent data sets.

Figure 2

Modelling the data statistics for a cubic insulin crystal. (a) The signal-to-noise ratio, , at a resolution h ⁻¹ = 1.5 Å for five sequential frames of Δ_ϕ = 1° and a |Φ| = 5° wedge of data versus the exposure time per frame. The calculations were carried out with (black line) and without (red line) accounting for radiation damage. (b) Signal-to-noise ratio and graphical solution of a set of equations (§2.2.2) at a resolution of 1.5 Å for three progressive subwedges. The signal-to-noise target was C = 2, Δ_ϕ was fixed at 1° and |Φ| = 5°. No solution was possible for the third subwedge. (c) The same as (b) for four consecutive wedges and resolution 1.55 Å.

Figure 3

Test-data collection for cubic insulin crystals. (a) Predicted and experimental R _merge versus resolution. (b) Predicted and experimental 〈J/σ(J)〉 versus resolution. (c) Predicted and experimental relative diffraction intensities, .

Figure 4

Test-data collection for the P19–siRNA-1 crystal. (a) Predicted and experimental R _merge (solid line) and (dashed line) versus resolution for P19–siRNA-1A (blue) and P19–siRNA-1B (red). (b) Predicted and experimental relative diffraction intensities, , versus the dose and resolution for P19–siRNA-1B.

Figure 5

Experimental R _merge (solid line) and (dashed line) versus resolution for test-data sets P19–siRNA-2A (black squares), P19–siRNA-2B (red triangles), P19–siRNA-2C (green circles) and P19–siRNA-2D (blue diamonds).

Figure 6

Test-data collection from an FAE crystal. (a) Predicted and experimental R _merge and versus resolution. (b) Predicted and experimental relative diffraction intensity, , versus dose and resolution. The nominal dose rate was 60 kGy s⁻¹ (see text for details); this rate multiplied by the cumulative exposure time is used as the dose axis.

Figure 7

Data collection from an FtsH crystal. (a) Predicted and experimental R _merge and ratio versus resolution. (b) Predicted and experimental relative diffraction intensity, , versus dose and resolution.