Small-angle X-ray scattering (SAXS) has emerged as a powerful and versatile tool in the structural biologist’s arsenal, but also one requiring careful interpretation1. Here, we highlight the danger of over-interpretation of SAXS data in ensemble modeling through a case study, the classical nuclear import receptor, importin-β2,3. This protein possesses a solenoid architecture that permits a spring-like extension, and this flexibility can be approximated by a single coordinate; the radius of gyration (Rg) 4. We exploit this to examine both the behavior of the protein and the robustness of ensemble modeling from SAXS, molecular dynamics and crystal structures.
We find that scattering data for apo importin-β can be described by an ensemble of conformers dominated by structures with Rgs within 3 Å of that of the RanGTP-bound structure. However, through restriction of the conformational pool used in selection, we also show that the scattering data is unable to differentiate between highly divergent ensembles; in fact, mutually-exclusive conformational pools can be identified that each lead to consistent restorations of the data. We demonstrate that ensemble fitting algorithms are able to adjust for large perturbations in the input models, and suggest that the identification of a dataset's tolerance for varying models is a key step in analysis.