Quality Control

This is an attempt of putting together a number of datasets with different characteristics (high and low resolution, good and bad crystals, untwinned and twinned), their evaluation with different versions of XDS up to the structure solution (as far as that can be done automatically), and the determination of the quality of the resulting data using experimental phasing and refinement.

The goal of this is to find generally optimal parameters for XDS data reduction, to compare new versions of XDS with older ones, and to discover bugs or regressions. At the same time, it serves to display standard procedures for solving crystal structures, which may be used for teaching or learning crystallography.

The metrics I would like to look at is

• anomalous signal: SHELXC anomalous correlations (internally or between wavelengths); SHELXD (fixed version) success rates and CCall/CCweak maximum values. Clemens Vonrhein additionally suggested anomalous difference fourier peak heights.
• refinement of a given (known) set of heavy atom positions in (e.g.) SHARP (fixed version): r.m.s. or absolute phase difference, or map correlation w.r.t. phases from good model
• refinement with (e.g.) refmac5 (fixed version) of a given model, finding LL and R/R_free

For each project mentioned below, both the raw data and the XDS data reduction is available - links are on the project pages, which are named according to their PDB ids.

The raw data (images) for the different datasets are either available from this site by FTP, or from the (publicly accessible!) JCSG dataset archive, or - for 1RQW - from [1] or [2].

There's currently data available for

• 2-wl MAD: PDB id 1T92 (PulG, 116 residues, 2 copies/ASU, spacegroup P6₅22, resolution 2.8 Å)
• 3-wl MAD: PDB id 1ZTV (JCSG target name TB1631F, resolution 3.1 Å) - evaluated with Qingping Xu; it's one of the JCSG datasets.
• native data: PDB id 2GIF (AcrB, spacegroup C2, resolution 2.9 Å - a large membrane protein structure)
• native data for MR: PDB id 1YCE (ATPase C-ring, spacegroup P2₁, resolution 2.4 Å - a large membrane protein structure, 44-fold NCS, strong diffuse scattering)
• 2-wl Bromide MAD: PDB id 1RQW (Thaumatin, a sweet-tasting protein of 207 resides, spacegroup P4₁2₁2, resolution 1.8 Å, collected at the DGK Workshop 2007; could also treat either peak or inflection as SAD. These datasets have been made available by Manfred S. Weiss and Annette Faust (see Xray Tutorial).

As part of the 2011 ACA workshop on data processing, organized by Ed Collins and Andy Torelli, I extensively document XDS data processing (and in most cases also structure solution) of

• 2VB1: hen egg-white lysozyme @ 0.65Å resolution, PDB id 2vb1. Data (sweeps a to h, each comprising 60 to 360 frames of 72MB) were collected by Zbigniew Dauter at APS 19-ID and are available from here. Details of data collection, processing and refinement are published.
• simulated-1g1c: James Holton's simulated (using his MLFSOM program) data of PDB id 1g1c - 100 synthetic datasets (15 frames each) strongly affected by radiation damage. These 100 datasets were simulated in random orientations and with random (?) "crystal" sizes. The processing and scaling can be considered non-standard, and challenging.
• 2QVO: S-SAD: PDB id AF1382, a 95-residue protein used by James Tucker Swindell II to establish optimized procedures for data reduction. The data available to solve the structure are two runs of 360° collected at 1.9Å wavelength.
• 3CSL: 3-wl SeMet-MAD data PDB id 3CSL, a complex of a 22 stranded beta-barrel outer membrane protein (the ordered residues 112-865 harbour 9 SeMet), its 173-residue hemophore (1 SeMet), and heme. Datasets are at [3]. 2 complexes per ASU, useful data to 3.2Å, useful anomalous data to about 5Å. Challenging for humans, and too difficult for automatic methods of structure solution and model building.
• 1Y13: SAD with a twist that requires some detective work.

Recently I've written SIM_MX which should make parts of testing and development of XDS independant from real data.