Date Published: March 01, 2017
Publisher: International Union of Crystallography
Author(s): Tobias Krojer, Romain Talon, Nicholas Pearce, Patrick Collins, Alice Douangamath, Jose Brandao-Neto, Alexandre Dias, Brian Marsden, Frank von Delft.
XChemExplorer is a graphical workflow and data-management tool for the parallel determination of protein–ligand complexes. Its implementation, usage and application are described here.
Protein crystallography is a cornerstone of structure-based ligand design (SBLD; Blundell et al., 2002 ▸), providing the molecular details of how newly synthesized compounds interact with the protein of interest in order to drive the compound-design process. This requires iterative cycles of design and synthesis of compounds with improved binding characteristics, based on structural information of previous protein–ligand complexes and biophysical measurements (Carvalho et al., 2009 ▸). In the past, SBLD was the preserve of industry, but academic groups have lately started to pick up on this approach (Hole et al., 2013 ▸; Brem et al., 2016 ▸).
Protein–ligand structure determination may be an algorithmically simple and linear process, but it turns into a challenge when many samples are analysed as part of an SBLD project which can go on for months and years. Not only are there specific demands on project design, but there have to be mechanisms to capture progress and results. This becomes exponentially challenging as the numbers of data sets become large, unless it is explicitly supported by software. For example, compounds often do not bind to a protein and the reason for this is not always immediately clear. While the experiment ultimately remains unsuccessful, the reason for the failure may be unrelated to the compound, and become evident only when every step of that particular experiment is analysed. XCE records compound information, metadata, file locations, observations, outcome and status of the project in a database. It serves as a funnel which channels disparate data into a grid so that it becomes comprehensible for the scientist. The program does not try to make decisions for the crystallographer, but it provides the tools to evaluate each experiment, assign a result and (re)run calculations in batches. The recommended workflow (Fig. 1 ▸) is reflected by the design of the graphical user interface (Figs. 2–5): each tab in the interface represent a milestone in the workflow, and the processes necessary to achieve the respective objectives are bundled in the coloured boxes at the bottom of the panel. The main workflow is grouped into four sections (data processing, initial map calculation, hit identification and refinement) and it is possible to enter at any point, provided that the required data are present in a suitable format.
XChemExplorer is implemented in Python (http://www.python.org) and, like CCP4, uses the PyQt4 library to provide the graphical user interface functionality. It runs on any Linux or Mac OS X system that has CCP4 v.7.0 or higher installed. XCE uses SQLite (http://www.sqlite.org) as a relational database-management system. The RDKit library (http://www.rdkit.org) is used to create two-dimensional images of compounds. Since XCE is a workflow and data-management tool and not an algorithm, it makes use of existing software for different parts of the structure-determination process: xia2 (Winter, 2010 ▸) for data processing; DIMPLE for initial refinement and map calculation; AceDRG (Long et al., 2017 ▸) for generation of ligand coordinates and restraints; REFMAC (Murshudov et al., 2011 ▸) for refinement; PanDDA (Pearce et al., 2016 ▸) for hit identification; a variety of tools from PHENIX (Adams et al., 2010 ▸) for validation purposes; and Coot (Emsley et al., 2010 ▸) for model building. There is, however, no fundamental limitation to incorporating other software packages. Source code for XChemExplorer and installation instructions are available at http://tkrojer.github.io/XChemExplorer.
XCE has been extensively tested by supporting crystallographic fragment screening at Diamond Light Source. In the course of several iterations of bug fixes, the program has become one of the cornerstones of the facility: only once the program was available in its current stable form did it become possible to completely model fragment-screening projects promptly and efficiently.