FragMAX: the fragment-screening platform at the MAX IV Laboratory

Abstract

Advances in synchrotron storage rings and beamline automation have pushed data-collection rates to thousands of data sets per week. With this increase in throughput, massive projects such as in-crystal fragment screening have become accessible to a larger number of research groups. The quality of support offered at large-scale facilities allows medicinal chemistry-focused or biochemistry-focused groups to supplement their research with structural biology. Preparing the experiment, analysing multiple data sets and prospecting for interesting complexes of protein and fragments require, for both newcomers and experienced users, efficient management of the project and extensive computational power for data processing and structure refinement. Here, FragMAX, a new complete platform for fragment screening at the BioMAX beamline of the MAX IV Laboratory, is described. The ways in which users are assisted in X-ray-based fragment screenings and in which the fourth-generation storage ring available at the facility is best exploited are also described.

Keywords: BioMAX; FragMAX; drug discovery; fragment screening; high-throughput data analysis; protein crystallography; software; user facility.

Figure 1

A typical experiment at FragMAX consists of five modules: (1) preparation of the crystals, manipulations with fragments, soaking of crystals and their harvesting, (2) data collection at the BioMAX beamline, (3) multiplex data analyses supported by the High-Performance Computing (HPC) infrastructure of the MAX IV Laboratory, (4) automated ligand-search methods and (5) export tools with support for the OneDep group deposition method. FragMAX projects are managed by our web application, FragMAXapp, which handles experimental information and results simply and intuitively.

Figure 2

Sample-preparation overview. At FragMAX, crystallization plates are set up using the Mosquito with user-provided crystallization reagents and proteins. Crystal soaking can be performed using a liquid handler or manually, assisted by a Crystal Shifter instrument. Sample annotation is performed with the Crystal Shifter software together with FragMAXapp, which synchronizes with the ISPyB database for beamline operation.

Figure 3

Data collection at BioMAX. FragMAX benefits from the modern and efficient infrastructure at the beamline. The ISARA sample changer supports large experiments (up to 464 crystals per single load), with fast sample changing and a very high success rate for crystal mounting. Experiments are controlled by the MXCuBE3 web interface and data processing is handled by software deployed at the MAX IV High-Performance Computing infrastructure.

Figure 4

FragMAXlib: a fragment library for FragMAX. (a) Basic description of the library. The core set contains 96 compounds with an average size of 11 non-H (heavy) atoms. Predicted atom-based partition coefficients SlogP were quantified according to Wildman & Crippen (1999 ▸). (b) Examples of library entries. Selection of the fragments was biased towards compounds with a close spatial arrangement of hydrogen-bond donors (D) and acceptors (A), i.e. with a topological pathway of one or two covalent bonds.

Figure 5

Evaluation of FragMAX performance. (a) Close-up view of the binding pocket for a hit found with FragMAXapp in a screening experiment with human carbonic anhydrase II. No ligand-related electron-density peaks were observed in a (2m|F _o| − D|F _c|) electron-density map from an autorefined method (i.e. direct output from DIMPLE, FSpipeline or BUSTER) contoured at 0.2973/1.1370 e Å⁻³ (1.50 r.m.s.d.). In contrast, PanDDA analysis and the corresponding event map revealed the ligand, and its relevance can be assessed by comparing the event map with the average map created with apo data sets. The figure shows the event map and the average map side by side, as available in FragMAXapp. Bottom: structure of the hit compound (left) and radial plot for the ligand-fit metrics (right) generated using the PanDDA giant scripts. (b) Comparison of the numbers of hits found with different proteins employing three different data-analysis approaches: a manual examination of unexplained density peaks found by Coot (‘Classic’, green bars), PanDDA analysis of a single combination of data-processing and structure-refinement pipelines (‘Standard PanDDA’, grey) and PanDDA analysis of the best solutions, defined by highest resolution, lowest R factors and highest ISa, from 18 possible combinations of data-analysis and structure-refinement pipelines offered in FragMAXapp (‘FragMAXapp’, orange). Endothiapepsin (hit rate = 30%) and Aar/RNaseH (hit rate = 21%) are user project targets; hCAII, human carbonic anhydrase II (hit rate = 15%); PROK, proteinase K (hit rate = 13%). The described hit rate is based on FragMAXapp analysis.