Research map / Probabilistic and scientific modeling / Molecules and proteins
Learning Protein Structure-Function Relationships through Knowledge-guided Representation Decomposition
Mingqing Wang, Zhiwei Nie, ATHANASIOS V. VASILAKOS, Yonghong He, Zhixiang Ren
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 |
|---|---|---|---|
| use ProtDiS knowledge decomposition instead of ESM3 structural tokenizer | ESM-3 structural embeddings · EC (structure-based split) | accuracy: 78.7 -> 83.5 | Improved |
| use ProtDiS instead of ESM3ST | ESM-3 structural embeddings · Ligand Affinity (structure-based split) | spearman: 35.1 -> 36.6 | Improved |
| use ProtDiS instead of ESM3ST | ESM-3 structural embeddings · SCOP-fa (structure-based split) | accuracy: 75.0 -> 78.0 | Improved |
| use ProtDiS instead of ESM3ST | ESM-3 structural embeddings · EC (random split) | accuracy: 88.2 -> 89.0 | Improved |
| use ProtDiS instead of ESM3ST | ESM-3 structural embeddings · MF, BP, CC, SCOP-cf (random split) | fmax/accuracy: MF 77.5 -> 77.2; BP 60.0 -> 60.0; SCOP-cf 80.9 -> 80.9 | No clear effect |
| use ProtDiS instead of ESM3ST | ESM-3 structural embeddings · Structure Similarity (random split) | spearman: 73.0 -> 72.4 | Hurt |
| use ProtDiS knowledge embeddings instead of ESM-3 structural embeddings with XGBoost | ESM-3 structural embeddings · enzyme function pair classification, highest TM-score bin | AUC: 0.868 -> 0.946 | Improved |
| use ProtDiS instead of ESM3ST | ESM-3 structural embeddings · PPIs (structure-based split) | auroc: 82.1 -> 84.6 | Improved |
Closest papers by meaning
- ProSST: Protein Language Modeling with Quantized Structure and Disentangled Attention (NeurIPS 2024 · similarity 0.89)
- ProtGO: Function-Guided Protein Modeling for Unified Representation Learning (NeurIPS 2024 · similarity 0.88)
- ProtCLIP: Function-Informed Protein Multi-Modal Learning (AAAI 2025 · similarity 0.88)
- Prot2Text-V2: Protein Function Prediction with Multimodal Contrastive Alignment (NeurIPS 2025 · similarity 0.87)
- Protein Design with Dynamic Protein Vocabulary (NeurIPS 2025 · similarity 0.87)
- Prot2Text: Multimodal Protein’s Function Generation with GNNs and Transformers (AAAI 2024 · similarity 0.86)
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.
