Hyperbolic stratification of protein intrinsic disorder and structure-mediated interactions in the human protein interactome

Classical models of protein-protein interactions (PPIs) focus on stable, structure-driven interfaces between folded domains, yet recent work highlights the central role of intrinsic disorder and phase separation in shaping dynamic, multivalent associations. How these interaction modes are reflected in the large-scale organization of PPI networks remains unclear.

Here, we map the human interactome onto a hyperbolic representation, integrating sequence- and structure-derived features to test whether network organization reflects distinct molecular interaction strategies. Radial position defines a continuum: central proteins are enriched in folded domains, structural complexity, and post-translational modifications, whereas peripheral proteins show increased intrinsic disorder and liquid-liquid phase separation (LLPS) propensity. Angular organization further reveals communities structured by characteristic domain architectures or disorder-linked motifs.

Combined analysis of intrinsic disorder, LLPS propensity, and binding-mode diversity uncovers interaction patterns associated with distinct molecular functions and motif repertoires. Condensate-associated proteins span multiple communities while retaining shared short linear motif signatures. Together, these results show that the hyperbolic map links sequence composition, structural organization, and network topology, providing a framework to interpret protein interaction behavior and to guide functional analysis within the human interactome.