Research map / Graph learning / Knowledge and temporal graphs
Targeted Pathway Inference for Biological Knowledge Bases via Graph Learning and Explanation
Rikuto Kotoge, Ziwei Yang, Zheng Chen, Yushun Dong, Yasuko Matsubara, Jimeng Sun, Yasushi Sakurai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add ESM-2 amino-acid sequence node features (ablation) | PathMamba · KEGG bio-network classification | overall accuracy: 0.440 -> 0.744 | Improved |
| scale ESM-2 encoder from 35M to 650M (ablation) | PathMamba · KEGG bio-network classification | overall accuracy: 0.722 -> 0.744 | Improved |
| scale ESM-2 encoder from 650M to 3B (ablation) | PathMamba · KEGG bio-network classification | overall accuracy: 0.744 -> 0.742 | No clear effect |
| use PathMamba instead of GPS as classifier | ExPath classifier · KEGG bio-network classification | overall accuracy: 0.726 -> 0.744 | Improved |
| add pathway-wise Mamba to GIN (PathMamba vs GIN) | GIN · KEGG bio-network classification | overall accuracy: 0.717 -> 0.744 | Improved |
| use PathMamba instead of GPS for efficiency | hybrid graph models · KEGG (batch size 32) | training time (msec): 29.2 -> 24.4 | Improved |
| use PathExplainer instead of GNNExplainer on PathMamba (ablation) | ExPath · KEGG pathway preservation | max path length: 9 -> 16 | Improved |
| use PathMamba instead of GPS under PathExplainer (ablation) | ExPath · KEGG pathway preservation | max path length: 12 -> 16 | Improved |
| use PathExplainer instead of GNNExplainer on PathMamba (ablation) | ExPath · KEGG GO enrichment | #EBF: 11.89 -> 14.77 | Improved |
| use PathMamba instead of GIN under PathExplainer (ablation) | ExPath · KEGG GO enrichment | ECS: 0.69 -> 0.84 | Improved |
| use PathExplainer with GPS backbone instead of GNNExplainer (ablation) | GPS · KEGG GO enrichment | #EBF: 8.88 -> 8.26 | Hurt |
| use PathExplainer with GIN backbone instead of GNNExplainer (ablation) | GIN · KEGG GO enrichment | ECS: 0.59 -> 0.69 | Mixed |
Closest papers by meaning
- MOTIVE: A Drug-Target Interaction Graph For Inductive Link Prediction (NeurIPS 2024 · similarity 0.84)
- KINDLE: Knowledge-Guided Distillation for Prior-Free Gene Regulatory Network Inference (NeurIPS 2025 · similarity 0.84)
- OgBench: A Framework for Evaluating Graph Neural Networks on Omics Data (NeurIPS 2026 · similarity 0.84)
- GeSubNet: Gene Interaction Inference for Disease Subtype Network Generation (ICLR 2025 · similarity 0.84)
- MAGE: Model-Level Graph Neural Networks Explanations via Motif-based Graph Generation (ICLR 2025 · similarity 0.84)
- Pre-Training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information Bottleneck (AAAI 2025 · similarity 0.83)
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.
