ARF-GNN: Adaptive receptive field graph neural network for protein function prediction
Zhiqiang Hui1, Weizhong Lu1, Yiyi Xia2
1School of Electronic and Information Engineering, Suzhou University of Science and Technology, Suzhou, 215009, China.
Computational Biology and Chemistry
|August 5, 2026
Summary
We developed ARF-GNN, a novel graph neural network for protein function prediction. It adaptively adjusts its receptive field to better model protein structures, significantly improving prediction accuracy.
Area of Science:
- Bioinformatics
- Computational Biology
- Structural Biology
Background:
- Protein function prediction is crucial for understanding cellular mechanisms and drug discovery.
- Traditional Graph Neural Networks (GNNs) struggle to model both local and long-range protein structural features due to fixed receptive fields.
- Accurate modeling of diverse structural contexts is essential for effective protein function prediction.
Purpose of the Study:
- To introduce ARF-GNN, an adaptive receptive field graph neural network designed to overcome the limitations of traditional GNNs in protein function prediction.
- To develop a model that can dynamically integrate structural information by adapting its receptive field size.
- To improve the accuracy and robustness of protein function prediction by effectively capturing multi-scale structural patterns.
Main Methods:
- Proposed ARF-GNN, an adaptive receptive field graph neural network.
- Implemented hierarchical multi-hop neighborhood aggregation for dynamic structural context modeling.
- Introduced a dual-branch meta-learning framework with Task and Meta branches for multi-label annotation and adaptive receptive field learning.
- Enabled structure-aware, input-adaptive information integration.
Main Results:
- ARF-GNN demonstrated significant improvements over existing benchmark models on the PDBch dataset.
- Achieved enhanced performance in AUPR, Fmax, and Smin evaluation metrics.
- Ablation and interpretability analyses confirmed the model's ability to capture functionally relevant multi-scale structural patterns.
Conclusions:
- ARF-GNN provides a principled paradigm unifying expressive structural representation with data-driven neighborhood adaptation for protein function prediction.
- The adaptive mechanism robustly captures functionally relevant structural patterns at multiple scales.
- ARF-GNN establishes a new state-of-the-art for protein function prediction by effectively integrating local and long-range structural information.
Related Concept Videos
02:26Protein Networks
An organism can have thousands of different proteins, and these proteins must cooperate to ensure the health of an organism. Proteins bind to other proteins and form complexes to carry out their functions. Many proteins interact with multiple other proteins creating a complex network of protein interactions.
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
01:49Conserved Binding Sites
Many proteins’ biological role depends on their interactions with their ligands, small molecules that bind to specific locations on the protein known as ligand-binding sites. Ligand-binding sites are often conserved among homologous proteins as these sites are critical for protein function.
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally analyses the...
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally analyses the...
02:26Conservation of Protein Domains Over Different Proteins
Protein domains are small structurally independent units that are part of a single amino acid chain. Although these domains are often structurally independent, they may rely on synergistic effects to perform their functions as part of a larger protein. Protein domains may be conserved within the same organism, as well as across different organisms.
A limited set of protein domains often duplicate and recombine during evolution. These domains can be organized in different combinations to form...
A limited set of protein domains often duplicate and recombine during evolution. These domains can be organized in different combinations to form...