Year: 2026
Type: preprint
Abstract: Small-angle X-ray scattering (SAXS) is routinely used for the structural characterization of protein monomers in solution, and computational prediction methods such as AlphaFold now provide accurate atomic models for most proteins. When a predicted model and the experimental scattering curve disagree, several refinement strategies exist, but selecting the appropriate one remains difficult, particularly for users without SAXS modeling expertise. ATSASfold is a web server developed to help analyse the results of SAXS measurements on protein monomers and select appropriate refinement strategies. From a provided experimental SAXS curve, ATSASfold performs primary data analysis, gives a general assessment of macromolecules shape, size, possible aggregation state and data quality. An AlphaFold-generated model is fitted against the SAXS curve and the agreement with the experiment is evaluated. In cases where the fit shows significant discrepancies, a machine-learning classifier trained on SASBDB depositions assigns the case to one of four categories, suggesting the most relevant ATSAS-based strategy for model refinement: fixed model fitting, rigid-body modelling, normal-mode analysis or ensemble optimization. When needed, a graphical interface allows the user to adjust domains or flexible regions of the protein before submitting the structure to ATSAS online for longer fitting jobs. After the fitting is completed, ATSASfold provides a summary of the results. The workflow was tested on examples from the SASBDB open web-databank. These use cases demonstrate smooth operation of the web server and show how ATSASfold can simplify the transition from SAXS data collection to structure-based interpretation.
Authors: Timur Tropin, Arne Raasakka, Clement Blanchet
Institution: European Molecular Biology Laboratory
Language: English
Subfield: Structural Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7236221300
