A computational pipeline for generating SAXS-consistent atomistic protein ensembles

Abstract

Understanding protein behaviour in solution remains a central challenge in structural biology. High-resolution experimental techniques such as macromolecular crystallography and, more recently, cryogenic electron microscopy (cryo-EM) provide atomic detail, but often miss the conformational heterogeneity and dynamics that define function. Small-angle X-ray scattering (SAXS) reports directly on size, shape, and flexibility in solution, though at lower resolution. Molecular dynamics (MD) simulations complement SAXS with atomistic insight but are limited by sampling, inaccuracies in force fields, and the need for experimental validation. Integrating SAXS with MD offers a powerful route to refine structural models towards realistic solution ensembles. Furthermore, the rise of machine learning structure prediction methods has transformed access to atomistic protein models, yet these predictions, largely trained on crystallographic data, often struggle to describe dynamic solution states. Such models provide excellent starting points but frequently require refinement against experiment. To address this, the AutoMD-SAXS workflow was developed: an accessible framework that automates MD setup, execution, and SAXS-guided analysis. Designed for deployment at Beamline B21, Diamond Light Source, it forms part of a broader pipeline for solution structure refinement.

The workflow was applied across three case studies. For the Rift Valley fever virus glycoprotein, AutoMD-SAXS identified flexible domain motions critical for viral entry. For engineered IgG2 antibody fragments, the novel refinement algorithm Carbonara optimised conformations against SAXS data, producing viable simulation seeds for enhanced ensemble generation. Finally, an IgG3 antibody with an extended hinge exposed force-field limitations and the need for enhanced sampling. AutoMD-SAXS provides a robust, user-friendly platform for SAXS-driven ensemble refinement, contributing to the growing toolkit of integrative structural biology software.

More information

Identifiers

ORCID for Jonathan Essex: orcid.org/0000-0003-2639-2746

Catalogue record

Export record