Research map / Label-efficient and robust learning / Recognition and long-tail
VenusX: Unlocking Fine-Grained Functional Understanding of Proteins
Yang Tan, Wenrui Gou, Bozitao Zhong, Huiqun Yu, Liang Hong, Bingxin Zhou
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 |
|---|---|---|---|
| evaluate on cross-family split instead of mixed-family split (ablation) | best residue-level model · VenusX Res Act / BindI | AUPR: best AUPR drops 70-80% | Hurt |
| evaluate on cross-family split instead of mixed-family split (ablation) | best residue-level model · VenusX Res Dom | AUPR: drop <10% | Hurt |
| use sequence-structure model (SaProt-650M) instead of sequence-only ESM2-t33 for unseen families | SaProt-650M · VenusX Res cross-family Act | AUPR: 0.143 -> 0.185 | Improved |
| use sequence-structure model (SaProt-650M) instead of sequence-only for mixed-family Act | SaProt-650M · VenusX Res Act MF50 | AUPR: 0.852 -> 0.745 | Hurt |
| use sequence-structure model (SaProt-650M) instead of ESM2-t33 for fragments | SaProt-650M · VenusX Frag Act | accuracy: 0.814 -> 0.928 | Improved |
| use sequence-structure model (SaProt-650M) instead of ESM2-t33 for fragments | SaProt-650M · VenusX Frag Act | Macro-F1: 0.605 -> 0.825 | Improved |
| use structure-only GVP-GNN for fragment classification | GVP-GNN · VenusX Frag Act / Motif | Macro-F1: Act 0.906; Motif 0.370 | Mixed |
| use structural alignment (Foldseek) instead of sequence alignment (BLAST) | Foldseek · VenusX Pair Act F50 | AUC (%): 52.9 -> 96.1 | Improved |
| use encoder-decoder ProtT5 instead of sequence-only encoders | ProtT5 · VenusX Pair BindI F50 | AUC (%): 98.5 vs 7-20 points lower for ESM2/ProtBert | Improved |
Closest papers by meaning
- MMSite: A Multi-modal Framework for the Identification of Active Sites in Proteins (NeurIPS 2024 · similarity 0.86)
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.86)
- ProtCLIP: Function-Informed Protein Multi-Modal Learning (AAAI 2025 · similarity 0.85)
- From Mechanistic Interpretability to Mechanistic Biology: Training, Evaluating, and Interpreting Sparse Autoencoders on Protein Language Models (ICML 2025 · similarity 0.85)
- PPDiff: Diffusing in Hybrid Sequence-Structure Space for Protein-Protein Complex Design (ICML 2025 · similarity 0.85)
- Multi-Scale Representation Learning for Protein Fitness Prediction (NeurIPS 2024 · similarity 0.85)
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.
