FlexENN: A Graph Neural Network for Binding Energy Prediction of Globular and Intrinsically Disordered Proteins

Abstract

Intrinsically disordered proteins (IDPs) drive a large fraction of cellular signaling, transcription and assembly through interfaces that lack a single defined geometry, making the prediction of their binding energies beyond the reach of methods calibrated on the rigid complementarity of folded complexes. Here we introduce an adaptive message passing graph neural network, FlexENN, that predicts binding energies across the full structural spectrum, from folded domains to complexes with one disordered partner. The FlexENN architecture constructs a local interface graph in which each residue node integrates relative geometry, dynamic descriptors, and sequence information, allowing the network to infer conformational fuzziness from a single structure. Benchmarked across folded complexes and IDP systems, including tubulin-tubulin interfaces within microtubules (MTs) and MTs in complex with microtubule associated proteins (MAPs), with reference binding affinity values extracted from our experiments, the model retrieves physically meaningful energetic profiles for complexes where rigid-body methods fail. More broadly, this work delivers an accurate, structure-aware approach to predicting binding free energies for the broad class of disordered-mediated complexes formed by IDPs.