Protein structures are typically represented as single, static models, yet biological function is often encoded in conformational flexibility. The rapid expansion of structural resources such as the Protein Data Bank and large-scale predicted models (e.g. AlphaFold) creates an opportunity to systematically characterise protein flexibility, but existing approaches are either limited to specific datasets or rely on RMSD/RMSF measures that are not directly comparable across proteins. Here we present Fluctome, a structural-alphabet–based framework and web resource for quantifying and visualising local conformational flexibility across diverse protein structural ensembles at scale.
Fluctome encodes each protein structure into a 3Di structural alphabet, capturing local backbone geometry as a sequence of discrete states. For a given protein, multiple conformers—sourced from experimental ensembles or independently generated models—are aligned and converted to 3Di sequences. At each residue position, Fluctome computes a per-position fluctuation score based on the distribution of 3Di states, integrating both state frequencies and their similarity via a 3Di substitution matrix. This produces a normalised, model-agnostic flexibility profile that can be directly compared between proteins and across datasets. To contextualise these scores, Fluctome permits joint inspection with complementary metrics such as backbone RMSD, B-factors, and AlphaFold pLDDT.
We implemented Fluctome as an interactive web server that maps fluctuation scores onto 3D structures through a Mol*–based viewer, enabling intuitive exploration of flexible loops, hinge regions, and putative allosteric sites. Large-scale analyses across curated ensembles highlight consistent patterns: catalytic residues and buried cores tend to exhibit low 3Di fluctuation, whereas surface loops, termini, and ligand-binding regions often show elevated flexibility. Case studies further demonstrate how Fluctome profiles can (i) reconcile discrepancies between experimental and predicted models, and (ii) prioritise functionally relevant regions for mutagenesis or variant interpretation.
Together, Fluctome provides a unified, scalable framework for describing the “fluctuation landscape” of proteins, bridging static structures and dynamic behaviour, and offering a foundation for future integration with dynamics-aware prediction and design tools.