AlphaFold-Metainference: predicting structural ensembles of disordered proteins

This video summarizes the study “AlphaFold prediction of structural ensembles of disordered proteins” by Brotzakis, Zhang, Murtada & Vendruscolo (Nature Communications, 2025) (PMID: 39952928). The authors introduce AlphaFold-Metainference, a hybrid method that uses inter-residue distance predictions from AlphaFold as structural restraints in molecular dynamics simulations to generate structural ensembles of intrinsically disordered and partially disordered proteins. By combining deep learning predictions and physical sampling, the approach yields conformational ensembles that better match experimental small-angle X-ray scattering (SAXS) and NMR data than single static models. This method extends the utility of AlphaFold beyond folded proteins, offering a route to characterize the structural heterogeneity and dynamics of disordered systems.