Intrinsically disordered proteins (IDPs) and regions are biologically important but challenging to study, as they lack a single stable folded structure and rather populate heterogeneous conformational ensembles. Several approaches offer precious insights into their structural ensemble-activity relationship, such as nuclear magnetic resonance and small-angle X-ray scattering. However, these experiments provide only ensemble-averaged data that could not be easily translated into the underlying conformational distribution. To address this issue, we developed bAIes, a Bayesian framework that combines AlphaFold2 predictions with a random coil model, enabling efficient and accurate generation of structural ensembles consistent with both low- and high-resolution experimental data. We applied bAIes to the C-terminal disordered tail of the adhesion GPCR ADGRV1, where mutations in this domain are associated with Usher2 syndrome, a rare condition that leads to a double sensory handicap. While overall this fragment is disordered in solution, the bAIes structural ensembles captured the formation of transient secondary structures in three segments, showing improved agreement with experimental data compared to both the AlphaFold2 prediction and the ensemble generated with the random coil model. Despite their transient nature, these helices might play a key role in binding to other partners through a folding-upon-binding mechanism, a process that might be altered in the case of mutations or post-translational modifications.
Héritier et al. (Sun,) studied this question.