Structure-Guided Adaptive Propagation for Protein-Protein Interaction Site Prediction

2026-06-01

AI Summary

This work addresses the limitations of existing graph neural networks in predicting protein–protein interaction sites, which rely on fixed propagation mechanisms that fail to account for the heterogeneity of local residue geometric environments and often misclassify structurally similar non-interacting residues. To overcome this, the authors propose SGAP-PPIS, an equivariant graph neural network that extracts multi-scale geometric features and dynamically generates adaptive propagation coefficients for each residue. This enables a balanced trade-off between preserving geometry-aware local features and diffusing information across neighborhoods. The model incorporates geometry-guided adaptive propagation, multi-scale alignment constraints, and multi-step state representations, thereby transcending conventional fixed-propagation paradigms. Evaluated on the Test_60 benchmark, SGAP-PPIS achieves state-of-the-art performance, with ablation studies confirming the contribution of each core component.

Abstract

Accurate prediction of protein-protein interaction sites (PPIS) is essential for understanding cellular processes, disease mechanisms, and therapeutic target discovery. Graph-based deep learning has advanced PPIS prediction by incorporating residue-level structural context. However, most graph-based models still rely on fixed propagation schemes that treat all residues similarly, despite the structural and functional heterogeneity of protein interfaces. Such propagation may limit the ability to adapt information diffusion to local geometric environments, making it difficult to distinguish true interaction sites from structurally similar non-interacting neighbors. We present SGAP-PPIS, a structure-guided adaptive propagation model for PPIS prediction. Rather than using a fixed propagation mechanism, SGAP-PPIS leverages multi-scale geometric states from an equivariant graph neural network to generate residue-wise propagation coefficients. This design allows each residue to adaptively balance local feature preservation and neighborhood diffusion according to its geometric microenvironment. Experimental results show that SGAP-PPIS achieves competitive performance among the state-of-the-art methods on Test\_60. Ablation studies show that geometry-conditioned adaptive propagation, scale-aligned geometric guidance, and multi-step propagation-state representation jointly drive these improvements.

Problem

• protein-protein interaction site prediction
• graph-based deep learning
• adaptive propagation
• structural heterogeneity
• geometric microenvironment

Innovation

• adaptive propagation
• structure-guided
• equivariant graph neural network
• multi-scale geometric states
• protein-protein interaction site prediction