Graph Neural Networks for Protein-Protein Interactions -- A Short Survey

TL;DR

This survey addresses the problem of predicting protein-protein interactions (PPIs) by leveraging graph-structured representations of biological data. It categorizes graph-based methods into two main families: GNN/GCN and GAT/Graph-AE/Graph-BERT, and reviews representative models and datasets across these groups. The discussion covers datasets such as STRING, SHS27K, and SHS148K, and highlights contributions like improved generalization for novel proteins, long-distance dependency modeling, and residue-level interpretation. The work underscores current challenges—label scarcity, domain shift, and computational constraints—and outlines directions for integrating richer graph information and refining architectures to enhance predictive accuracy and robustness in PPI prediction.

Abstract

Protein-protein interactions (PPIs) play key roles in a broad range of biological processes. Numerous strategies have been proposed for predicting PPIs, and among them, graph-based methods have demonstrated promising outcomes owing to the inherent graph structure of PPI networks. This paper reviews various graph-based methodologies, and discusses their applications in PPI prediction. We classify these approaches into two primary groups based on their model structures. The first category employs Graph Neural Networks (GNN) or Graph Convolutional Networks (GCN), while the second category utilizes Graph Attention Networks (GAT), Graph Auto-Encoders and Graph-BERT. We highlight the distinctive methodologies of each approach in managing the graph-structured data inherent in PPI networks and anticipate future research directions in this domain.