Item

Abstract

The stunning success of AlphaFold2 (AF2) in the CASP14 protein structure prediction competition represented a watershed moment for structural biology. For NMR spectroscopy this has been seen an existential crisis, with AF2 models now widely regarded as more accurate than NMR structures, especially for the small proteins where NMR is usually applied [1]. However, it is now becoming clear that the two are complementary methods, with NMR able to provide data on flexible regions and proteins with distinct functional states where AF2 models may be less reliable [2,3]. We have previously described the CoMAND method, which uses a powerful R-factor expressing the match between experimental and back-calculated NOESY spectra to map protein conformational states [4]. Here we extend this to a hybrid method coupling the accuracy of AF2 in terms of the global fold to detailed descriptions of local micro-sates and their populations. An advantage of this method is the ability to select structural ensembles from unrestrained molecular dynamics trajectories, resulting in thermodynamically relevant ensembles, something that can not be achieved in conventional NMR structure determination based on restraints. I will present several examples of CoMAND analysis, including for the thalidomide binding domain of the E3 ubiquitin ligase cereblon, where CoMAND has been able to characterize large-scale changes in response to ligand binding, providing a model for the action of immunomodulatory drugs.