Research map / Vision-language and multimodal / Human and embodied modeling
PEER: A Comprehensive and Multi-Task Benchmark for Protein Sequence Understanding
Minghao Xu, Zuobai Zhang, Jiarui Lu, Zhaocheng Zhu, Yangtian Zhang, Chang Ma, Runcheng Liu, Jian Tang
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 ESM-1b instead of the strongest from-scratch encoder | Single-task protein sequence modeling · β-lactamase activity prediction | Spearman's ρ: 0.781 -> 0.839 | Improved |
| Use ESM-1b rather than the strongest from-scratch encoder | Single-task protein sequence modeling · Fold classification | accuracy: 10.93 -> 29.95 | Improved |
| Add contact prediction as an auxiliary task (ablation) | ESM-1b · β-lactamase activity prediction | Spearman's ρ: 0.839 -> 0.899 | Improved |
| Add secondary structure prediction as an auxiliary task (ablation) | CNN · Stability prediction | Spearman's ρ: 0.637 -> 0.695 | Improved |
| Add fold classification as an auxiliary task (ablation) | CNN · Stability prediction | Spearman's ρ: 0.637 -> 0.472 | Hurt |
| Add contact prediction as an auxiliary task (ablation) | Transformer · β-lactamase activity prediction | Spearman's ρ: 0.261 -> 0.142 | Hurt |
| Add fold classification as an auxiliary task (ablation) | Transformer · β-lactamase activity prediction | Spearman's ρ: 0.261 -> 0.276 | Improved |
| Add contact prediction as an auxiliary task (ablation) | ESM-1b · Contact prediction | L/5 precision: 45.78 -> - | Hurt |
| Add fold classification as an auxiliary task (ablation) | ESM-1b · Contact prediction | L/5 precision: 45.78 -> 35.86 | Hurt |
| Add secondary structure prediction as an auxiliary task (ablation) | ESM-1b · Contact prediction | L/5 precision: 45.78 -> 32.03 | Hurt |
| Increase sequence truncation length (ablation) | DDE, CNN, and ESM-1b · Subcellular localization and fold classification | performance | Improved |
| Increase training set size (ablation) | DDE, CNN, and ESM-1b · β-lactamase activity prediction | Spearman's ρ | Improved |
Closest papers by meaning
- Feature Reuse and Scaling: Understanding Transfer Learning with Protein Language Models (ICML 2024 · similarity 0.86)
- PRING: Rethinking Protein-Protein Interaction Prediction from Pairs to Graphs (NeurIPS 2025 · similarity 0.86)
- Prot2Text-V2: Protein Function Prediction with Multimodal Contrastive Alignment (NeurIPS 2025 · similarity 0.86)
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.86)
- Training Compute-Optimal Protein Language Models (NeurIPS 2024 · similarity 0.86)
- PPDiff: Diffusing in Hybrid Sequence-Structure Space for Protein-Protein Complex Design (ICML 2025 · 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.
