Research map / Label-efficient and robust learning / Self-supervised and contrastive
Evaluating Protein Transfer Learning with TAPE
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 |
|---|---|---|---|
| Add self-supervised pretraining (ablation) | LSTM · CB513 | accuracy: 0.71 -> 0.75 | Helped |
| Add self-supervised pretraining (ablation) | LSTM · ProteinNet CASP12 short-range contacts | P@L/5: 0.33 -> 0.49 | Helped |
| Add self-supervised pretraining (ablation) | LSTM · ProteinNet CASP12 medium-range contacts | P@L/5: 0.19 -> 0.34 | Helped |
| Add self-supervised pretraining (ablation) | LSTM · ProteinNet CASP12 long-range contacts | P@L/5: 0.15 -> 0.27 | Helped |
| Add self-supervised pretraining (ablation) | LSTM · Remote Homology Fold | Top-1 accuracy: 0.12 -> 0.26 | Helped |
| Add self-supervised pretraining (ablation) | Transformer · Fluorescence full test set | Spearman's ρ: 0.22 -> 0.68 | Helped |
| Add self-supervised pretraining (ablation) | Transformer · Stability | Spearman's ρ: -0.06 -> 0.73 | Helped |
| Use alignment features instead of learned pretrained features | Secondary structure prediction · CB513 | accuracy: 0.75 -> 0.80 | Helped |
| Use alignment features instead of learned pretrained features | Contact prediction · ProteinNet CASP12 short-range contacts | P@L/5: 0.49 -> 0.66 | Helped |
| Use alignment features instead of learned pretrained features | Remote homology detection · Remote Homology Fold | Top-1 accuracy: 0.26 -> 0.09 | Hurt |
| Use self-supervised pretraining with ResNet (ablation) | ResNet · Fluorescence full test set | Spearman's ρ: -0.28 -> 0.21 | Helped |
Closest papers by meaning
- Feature Reuse and Scaling: Understanding Transfer Learning with Protein Language Models (ICML 2024 · similarity 0.88)
- ProtGO: Function-Guided Protein Modeling for Unified Representation Learning (NeurIPS 2024 · similarity 0.86)
- Training Compute-Optimal Protein Language Models (NeurIPS 2024 · similarity 0.86)
- C3P: Contrastive promoter-protein pretraining yields representations capturing bacterial gene regulation (NeurIPS 2026 · similarity 0.85)
- Diffusion on Language Model Encodings for Protein Sequence Generation (ICML 2025 · similarity 0.84)
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.84)
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.