OGB
ogbn-proteins
Structure
1Graphs
132,534Nodes
39,561,252Edges
112Classes
8Node Feature Dim
8Edge Feature Dim
0.623–0.94Homophily
Coverage
108Results
1Variants
18Papers
- Latent-Hysteresis Graph ODEs: Modeling Coupled Topology-Feature Evolution via Continuous Phase Transitions Latent-Hysteresis Graph ODEs: Modeling Coupled Topology-Feature Evolution via Continuous Phase Transitions
- AdvSynGNN: Structure-Adaptive Graph Neural Nets via Adversarial Synthesis and Self-Corrective Propagation AdvSynGNN: Structure-Adaptive Graph Neural Nets via Adversarial Synthesis and Self-Corrective Propagation
- Quantum Graph Attention Network: A Novel Quantum Multi-Head Attention Mechanism for Graph Learning Quantum Graph Attention Network: A Novel Quantum Multi-Head Attention Mechanism for Graph Learning
- Efficient Mixed Precision Quantization in Graph Neural Networks Efficient Mixed Precision Quantization in Graph Neural Networks
- TMLR 2024: Extending Graph Condensation to Multi-Label Datasets: A Benchmark Study Extending Graph Condensation to Multi-Label Datasets: A Benchmark Study TMLR 2024
- Learning to Model Graph Structural Information on MLPs via Graph Structure Self-Contrasting Learning to Model Graph Structural Information on MLPs via Graph Structure Self-Contrasting
- ICLR 2024: Node Identifiers: Compact, Discrete Representations for Efficient Graph Learning Node Identifiers: Compact, Discrete Representations for Efficient Graph Learning ICLR 2024
- ICLR 2024: Graph Sparsification via Mixture of Graphs Graph Sparsification via Mixture of Graphs ICLR 2024
- Smoothed Graph Contrastive Learning via Seamless Proximity Integration Smoothed Graph Contrastive Learning via Seamless Proximity Integration
- Label Deconvolution for Node Representation Learning on Large-scale Attributed Graphs against Learning Bias Label Deconvolution for Node Representation Learning on Large-scale Attributed Graphs against Learning Bias
- IJCAI 2023: Hierarchical Transformer for Scalable Graph Learning Hierarchical Transformer for Scalable Graph Learning IJCAI 2023
- NeurIPS 2023: Graph Mixture of Experts: Learning on Large-Scale Graphs with Explicit Diversity Modeling Graph Mixture of Experts: Learning on Large-Scale Graphs with Explicit Diversity Modeling NeurIPS 2023
- CAP: Co-Adversarial Perturbation on Weights and Features for Improving Generalization of Graph Neural Networks CAP: Co-Adversarial Perturbation on Weights and Features for Improving Generalization of Graph Neural Networks
- SoGCN: Second-Order Graph Convolutional Networks SoGCN: Second-Order Graph Convolutional Networks
- ICML 2021: Training Graph Neural Networks with 1000 Layers Training Graph Neural Networks with 1000 Layers ICML 2021
- Residual Network and Embedding Usage: New Tricks of Node Classification with Graph Convolutional Networks Residual Network and Embedding Usage: New Tricks of Node Classification with Graph Convolutional Networks
- IJCAI 2020: Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification IJCAI 2020
- Understanding and Resolving Performance Degradation in Deep Graph Convolutional Networks Understanding and Resolving Performance Degradation in Deep Graph Convolutional Networks
112-label node classification Predict all applicable Gene Ontology-derived function labels for each protein. Edges have eight confidence values between zero and one, one per association evidence channel. Evaluated by ROC-AUC.
Dataset details
Task
Predict all applicable Gene Ontology-derived function labels for each protein. This is 112-task multi-label node classification evaluated by mean ROC-AUC across tasks.
Data
One undirected graph contains 132,534 proteins from eight species and biologically meaningful STRING associations including physical interaction, co-expression, and homology.
Features
Edges have eight confidence values between zero and one, one per association evidence channel. The standard OGB baseline derives node inputs by aggregating incident edge features; users must not assume that independent sequence features are supplied.
Splits and evaluation
Protein nodes are divided by species so validation and test require transfer to species not represented by training nodes. ROC-AUC is computed for each of the 112 function labels and then averaged.
Quirks and pitfalls
Associations are evidence-weighted rather than verified physical contacts, and confidence channels can reflect database coverage. Function annotation is incomplete, so unannotated functions may not be true biological negatives. Random node splitting breaks the intended cross-species test.
Sources
SOTA Progression
new SOTA
matches SOTA at that time
other proposed result
other baseline result
Leaderboard
| # | Model | ROC-AUC ↑ | Accuracy | F1-macro | F1-micro | Paper |
|---|---|---|---|---|---|---|
| 1 | LD | 0.8942
±
0.0007 | — | — | — | |
| 2 | X_WO/LD | 0.8915
±
0.0012 | — | — | — | |
| 3 | X_PNE | 0.8894
±
0.0014 | — | — | — | |
| 4 | X_GLEM | 0.8872 | — | — | — | |
| 5 | X_FNE | 0.8866
±
0.0008 | — | — | — | |
| 6 | RevGNN | 0.8824
±
0.0015 | — | — | — | |
| 7 | RevGNN-Deep | 0.8774
±
0.0013 | — | — | — | |
| 8 | UniMP+CEF | 0.8691
±
0.0018 | — | — | — | |
| 9 | GAT | 0.8682
±
0.0021 | — | — | — | |
| 10 | UniMP | 0.8642
±
0.0008 | — | — | — | |
| 11 | DeeperGCN | 0.8616
±
0.0016 | — | — | — | |
| 12 | DeeperGCN | 0.8580
±
0.0017 | — | — | — | |
| 13 | STR-Sparse | 0.8550 | — | — | — | |
| 14 | DeepGCN | 0.8496
±
0.0028 | — | — | — | |
| 15 | APPNP | 0.8450 | — | — | — | |
| 16 | MWE-DGCN | 0.8436
±
0.0065 | — | — | — | |
| 17 | GraphSAGE | 0.8430 | — | — | — | |
| 18 | GAT | 0.8410 | — | — | — | |
| 19 | GCN | 0.8320 | — | — | — | |
| 20 | GEN + FLAG + node2vec | 0.8251
±
0.0043 | — | — | — | |
| 21 | NodeNorm | 0.8226
±
0.0093 | — | — | — | |
| 22 | GEN | 0.8130
±
0.0065 | — | — | — | |
| 23 | GEN + FLAG | 0.8129
±
0.0067 | — | — | — | |
| 24 | HGODE | 0.8124
±
0.0063 | — | — | — | |
| 25 | SS/RD | 0.8120 | — | — | — | |
| 26 | DRAGON | 0.8046
±
0.0042 | — | — | — | |
| 27 | FROND | 0.8026
±
0.0049 | — | — | — | |
| 28 | DAGNN | 0.7995 | — | — | — | |
| 29 | SGFormer | 0.7953
±
0.0038 | — | — | — | |
| 30 | GATv2 | 0.7952
±
0.0055 | — | — | — | |
| 31 | SAGE | 0.7943
±
0.0075 | — | — | — | |
| 32 | Polynormer | 0.7941
±
0.0056 | — | — | — | |
| 33 | QGAT | 0.7941
±
0.0021 | — | — | — | |
| 34 | GEN | 0.7936
±
0.0086 | — | — | — | |
| 35 | SoGCN | 0.7928
±
0.0047 | — | — | — | |
| 36 | FLODE | 0.7923
±
0.0075 | — | — | — | |
| 37 | GREAD-BS | 0.7921
±
0.0074 | — | — | — | |
| 38 | Sheaf Diff. | 0.7896
±
0.0056 | — | — | — | |
| 39 | 4th-Order GCN | 0.7895
±
0.0057 | — | — | — | |
| 40 | BuNN | 0.7892
±
0.0048 | — | — | — | |
| 41 | GAT | 0.7863
±
0.0162 | — | — | — | |
| 42 | 6th-Order GCN | 0.7861
±
0.0042 | — | — | — | |
| 43 | GREAD-Exp | 0.7861
±
0.0068 | — | — | — | |
| 44 | Graph-MLP | 0.7850 | — | — | — | |
| 45 | GeniePath-BS | 0.7825
±
0.0035 | — | — | — | |
| 46 | HSGT | 0.7813
±
0.0025 | — | — | — | |
| 47 | GaAN | 0.7803
±
0.0073 | — | — | — | |
| 48 | GCN | 0.7784
±
0.0056 | — | — | — | |
| 49 | MoG | 0.7778
±
0.0038 | — | — | — | |
| 50 | GraphSAGE | 0.7768
±
0.0020 | — | — | — |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time