GeoGraph: Geometric and Graph-based Ensemble Descriptors for Intrinsically Disordered Proteins

TL;DR

GeoGraph tackles the challenge of describing intrinsically disordered protein ensembles by learning from coarse-grained MD contact graphs to predict ensemble-averaged, residue-level descriptors directly from sequence. The method uses a transformer-based sequence-to-sequence model with separate heads for geometric and graph-based targets, trained on CALVADOS-2 ensembles and evaluated on the Human–IDRome dataset. It achieves state-of-the-art or competitive performance on complex shape descriptors while offering vastly faster inference than full ensemble simulations, and its graph-derived representations prove highly transferable over PLM embeddings. The work also provides ablations and comparisons to STARLING and ALBATROSS, illustrating the benefits of simulation-informed, residue-level, graph-aware learning for IDPs. Public releases of GeoGraph and related resources enable broader adoption and further development in the IDP modeling community.

Abstract

While deep learning has revolutionized the prediction of rigid protein structures, modelling the conformational ensembles of Intrinsically Disordered Proteins (IDPs) remains a key frontier. Current AI paradigms present a trade-off: Protein Language Models (PLMs) capture evolutionary statistics but lack explicit physical grounding, while generative models trained to model full ensembles are computationally expensive. In this work we critically assess these limits and propose a path forward. We introduce GeoGraph, a simulation-informed surrogate trained to predict ensemble-averaged statistics of residue-residue contact-map topology directly from sequence. By featurizing coarse-grained molecular dynamics simulations into residue- and sequence-level graph descriptors, we create a robust and information-rich learning target. Our evaluation demonstrates that this approach yields representations that are more predictive of key biophysical properties than existing methods.