Research map / Graph learning / Graph self-supervised learning
Motif-based Graph Self-Supervised Learning for Molecular Property Prediction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Pre-train GIN with motif generation in BFS order | GIN · MUV | ROC-AUC (%): 78.7±1.5 vs 75.8±1.7 | Helped |
| Pre-train GIN with motif generation in BFS order | GIN · ClinTox | ROC-AUC (%): 80.7±2.1 vs 77.5±3.1 | Helped |
| Pre-train GIN with motif generation in BFS order | GIN · SIDER | ROC-AUC (%): 61.8±0.8 vs 60.7±0.5 | Helped |
| Pre-train GIN with motif generation in BFS order | GIN · HIV | ROC-AUC (%): 78.8±1.2 vs 77.8±1.4 | Helped |
| Pre-train GIN with motif generation in BFS order | GIN · Tox21 | ROC-AUC (%): 76.5±0.3 vs 77.2±0.4 | Hurt |
| Pre-train GIN with motif generation in BFS order | GIN · BACE | ROC-AUC (%): 79.1±0.9 vs 79.7±0.8 | Hurt |
| Pre-train GIN with motif generation in BFS order | GIN · ToxCast | ROC-AUC (%): 64.1±0.7 vs 63.5±0.4 | Helped |
| Pre-train GIN with motif generation in BFS order | GIN · BBBP | ROC-AUC (%): 69.7±0.9 vs 70.5±1.1 | Hurt |
| Use BFS rather than DFS motif generation (ablation) | MGSSL · MoleculeNet 8-dataset average | ROC-AUC (%): 73.5 vs 73.7 | Helped |
| Add atom-level pre-training tasks to motif-level pre-training (ablation) | MGSSL · MoleculeNet 8-dataset average | ROC-AUC (%): 73.0 -> 73.7 | Helped |
| Use unified multi-level pre-training instead of sequential pre-training (ablation) | MGSSL · MoleculeNet 8-dataset average | ROC-AUC (%): 73.4 -> 73.7 | Helped |
| Use MGSSL pre-training | GCN, GIN, RGCN, DAGNN, and GraphSAGE · MoleculeNet 8-dataset average | ROC-AUC (%): 68.8 -> 72.7; 67.0 -> 73.7; 68.3 -> 73.0; 67.1 -> 72.3; 68.3 -> 73.4 | Helped |
Closest papers by meaning
- MAGE: Model-Level Graph Neural Networks Explanations via Motif-based Graph Generation (ICLR 2025 · similarity 0.90)
- Towards Multiscale Graph-based Protein Learning with Geometric Secondary Structural Motifs (NeurIPS 2025 · similarity 0.90)
- Pre-Training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information Bottleneck (AAAI 2025 · similarity 0.89)
- On the Scalability of GNNs for Molecular Graphs (NeurIPS 2024 · similarity 0.89)
- Graph Generative Pre-trained Transformer (ICML 2025 · similarity 0.88)
- DMol: A Highly Efficient and Chemical Motif-Preserving Molecule Generation Platform (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.