New research by Z.Faidon Brotzakis from Skretas Lab concerning the structural dynamics of disordered proteins.
Dr Z. Faidon Brotzakis, senior post-doctoral researcher at the Skretas lab, in collaboration with Shengyu Zhang, Mhd. Hussein Murtada and Prof. Michele Vendruscolo from the Department of Chemistry of the University of Cambridge, recently published in Nature Communications a research article titled “AlphaFold prediction of structural ensembles of disordered proteins”. Their study describes a multidisciplinary approach combining AlphaFold data and Molecular Dynamics to characterize the structural dynamics disordered proteins.
Approximately 30% of the human proteome consists of intrinsically disordered proteins (IDPs) or proteins with intrinsically disordered regions (IDRs), which challenge conventional experimental and computational structure prediction methods. These regions fail to adopt a stable secondary or tertiary structure and undergo rapid conformational changes, making their characterization difficult. Despite their lack of a defined structure, IDRs play critical roles in cellular processes, including molecular recognition, signalling, and regulation. Their flexibility allows them to interact with multiple binding partners, contributing to dynamic cellular functions and adaptation. However, also often predisposes them to pathological misfolding and aggregation, contributing to neurodegenerative and other protein-misfolding diseases. Historically, their structural dynamics have remained largely unresolved, making them a "dark" and elusive part of the proteome.
This study developed and benchmarked the AlphaFold-Metainference algorithm, which integrates machine learning predictions from AlphaFold with molecular dynamics simulations, to determine the conformational ensembles of intrinsically and partially disordered proteins. This approach provides insights into the transient interactions of disordered proteins, offering a more comprehensive understanding of their structural dynamics.
The researchers demonstrated that AlphaFold-Metainference enhances structure prediction for disordered proteins compared to AlphaFold2 alone and achieves similar or higher accuracy in 80% of cases tested compared to conventional molecular dynamics when benchmarked against SAXS data. These findings illustrate how integrating molecular simulations with machine learning predictions enables a more detailed characterization of disordered proteins, contributing to the broader understanding of their role in health and disease.
https://www.nature.com/articles/s41467-025-56572-9
