Résumé
X-ray crystallography and NMR can provide detailed structural information of protein–protein complexes, but technical problems make their application challenging in the high-throughput regime. Other methods such as small-angle X-ray scattering (SAXS) are more promising for large-scale application, but at the cost of lower resolution, which is a problem that can be solved by complementing SAXS data with theoretical simulations. Here, we propose a novel strategy that combines SAXS data and accurate protein–protein docking simulations. The approach has been benchmarked on a large pool of known structures with synthetic SAXS data, and on three experimental examples. The combined approach (pyDockSAXS) provided a significantly better success rate (43% for the top 10 predictions) than either of the two methods alone. Further analysis of the influence of different docking parameters made it possible to increase the success rates for specific cases, and to define guidelines for improving the data-driven protein–protein docking protocols.
[Display omitted]
► pyDockSAXS: a new method for modeling protein-protein complexes by SAXS and docking. ► Binding energy can drive and complement model building based on SAXS data. ► The combined approach provided much better success rates than docking or SAXS alone. ► Large variety of protein assemblies can be accurately modelled with pyDockSAXS.