ATSAS 2.8: a comprehensive data analysis suite for small-angle scattering from macromolecular solutions

Abstract

ATSAS is a comprehensive software suite for the analysis of small-angle scattering data from dilute solutions of biological macromolecules or nanoparticles. It contains applications for primary data processing and assessment, ab initio bead modelling, and model validation, as well as methods for the analysis of flexibility and mixtures. In addition, approaches are supported that utilize information from X-ray crystallography, nuclear magnetic resonance spectroscopy or atomistic homology modelling to construct hybrid models based on the scattering data. This article summarizes the progress made during the 2.5-2.8 ATSAS release series and highlights the latest developments. These include AMBIMETER, an assessment of the reconstruction ambiguity of experimental data; DATCLASS, a multiclass shape classification based on experimental data; SASRES, for estimating the resolution of ab initio model reconstructions; CHROMIXS, a convenient interface to analyse in-line size exclusion chromatography data; SHANUM, to evaluate the useful angular range in measured data; SREFLEX, to refine available high-resolution models using normal mode analysis; SUPALM for a rapid superposition of low- and high-resolution models; and SASPy, the ATSAS plugin for interactive modelling in PyMOL. All these features and other improvements are included in the ATSAS release 2.8, freely available for academic users from https://www.embl-hamburg.de/biosaxs/software.html.

Keywords: ATSAS; biological macromolecules; data analysis; small-angle scattering; structural modelling.

Figure 1

Publications referencing ATSAS (1) out of all publications using biological solution scattering (2; left axis). Download numbers of the ATSAS package since 2006 (3; right axis).

Figure 2

Overall scheme of the ATSAS application suite.

Figure 3

Overview of SAS-based ab initio modelling. The modelling process starts by selection of a suitable data range and determination of the corresponding p(r) function. The chosen ab initio modelling approach is repeated several times in order to explore the available solution space, and the generated models are grouped according to their structural similarity. The most probable and average models are selected on the basis of the clustering. Pairwise FSC functions of the structurally aligned bead or dummy residue models are computed. The average of all pairwise FSC functions is used to determine the variability estimate Δ_ens as 2π/s _ens, where s _ens is the momentum transfer value at which the average FSC drops below 0.5. The corresponding resolution is estimated from the variability.

Figure 4

SREFLEX, flexible refinement of high-resolution models based on SAXS and normal mode analysis. Vectors show the conformational change modelled for hepatitis C virus NS3 helicase by SREFLEX when starting from PDB code 8ohm (Cho et al., 1998 ▸) (unbound conformation, blue) guided by the SAXS profile of a nucleotide-bound conformation (dots, simulated from PDB code 3kqn; Gu & Rice, 2010 ▸).

Figure 5

Ensemble optimization method (EOM). R _g parameter distributions for wild type (WT, upper panel) and a disulphide-stabilized mutant (MUT, lower panel) of urokinase plasminogen activator protein (SASBDB IDs SASDAT4 and SASDAU4, respectively; Mertens et al., 2012 ▸). The distributions of a pool of 10 000 randomized conformations, preserving individual domain structure, are shown as broken lines. The distributions of optimized ensembles selected by the genetic algorithm are shown as blue (WT) and red (MUT) bars, respectively. The decreased width of the distribution of selected structures for mutant relative to wild type indicates a reduction in flexibility, and the observed shift to smaller R _g values provides evidence of structural compaction. The metrics R _flex and R _σ are calculated from the distributions as 82% and 1.0 (wild type), and 45% and 0.1 (mutant). The R _flex value of the random pool is calculated as 85%.

Figure 6

PRIMUS/qt, the cross-platform SAS data analysis platform of ATSAS, providing (a) the main window with (1) a plot area for 1000+ simultaneous datasets, (2) advanced zoom capabilities, (3) advanced file filtering and selection, (4) direct file manipulation, (5) information about the selected file, (6) easily accessible analysis, and (7) data processing options. A variety of analysis wizards are implemented as frontends for convenient and reliable manual analysis of SAS data, employing the various ATSAS applications in the background. So far are available (b) the Guinier Wizard to determine the radius of gyration, (c) the Distance Distribution Wizard to determine the maximum dimension, (d) the Porod Wizard to determine the Porod volume, (e) the Shape Wizard for ab initio shape determination, including averaging and refinement, (f) the CRYSOL wizard to compute the fit of a priori models, (g) the OLIGOMER wizard for analysis of mixtures with known components, and (h) the Singular Value Decomposition Wizard for mixture analysis with unknown components.

Figure 7

POLYSAS GUI for SAXS data modelling of hNGF concentration dependence in solution using an oligomeric mixture of dimers and dimers of dimers (Covaceuszach et al., 2015 ▸).

Figure 8

Example of SASpy workflow, where structural models can be modified and refined while interactively evaluating their fit to SAXS experimental data.

Figure 9

A CHROMIXS screenshot displaying the plot (blue) of integrated intensities versus time (frame number) for a SEC-SAXS run at the EMBL P12 BioSAXS beamline (Blanchet et al., 2015 ▸). The user has selected a sample region (green) and a buffer region (red) has been predicted automatically by CHROMIXS.