Research map / Label-efficient and robust learning / Self-supervised and contrastive
Protein Fold Classification at Scale: Benchmarking and Pretraining
Dexiong Chen, Andrei Manolache, Mathias Niepert, Karsten Borgwardt
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 masking at ratio 0.9 instead of 0.0 (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 45.70 -> 58.52 | Improved |
| add masking at ratio 0.9 instead of 0.0 (ablation) | MiAE-B · TEDBench test, fine-tuning | macro F1: 71.41 -> 73.71 | Improved |
| add inverse folding loss (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 52.55 -> 58.52 | Improved |
| incorporate amino acid sequence (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 58.52 -> 62.14 | Improved |
| use span masking instead of random masking (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 58.52 -> 59.23 | No clear effect |
| increase decoder depth from 2 to 4 (mean pooling) (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 58.52 -> 59.65 | Improved |
| increase decoder depth from 2 to 4 (CLS pooling) (ablation) | MiAE-B · TEDBench test, linear probing CLS | macro F1: 34.74 -> 13.26 | Hurt |
| widen decoder from 512 to 768 (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 58.52 -> 27.83 | Hurt |
| narrow decoder from 512 to 256 (ablation) | MiAE-B · TEDBench test, linear probing | macro F1: 58.52 -> 35.50 | Hurt |
| scale encoder from MiAE-S to MiAE-L (ablation) | MiAE · TEDBench test, linear probing | macro F1: 49.43 -> 63.50 | Improved |
| scale encoder from MiAE-S to MiAE-L (ablation) | MiAE · TEDBench test, fine-tuned | macro F1: 72.28 -> 73.47 | No clear effect |
| fine-tune end to end instead of linear probing | MiAE-B+seq · TEDBench test | macro F1: 62.14 -> 74.56 | Improved |
Closest papers by meaning
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.89)
- FoldToken: Learning Protein Language via Vector Quantization and Beyond (AAAI 2025 · similarity 0.87)
- AlphaFold Database Debiasing for Robust Inverse Folding (NeurIPS 2025 · similarity 0.87)
- Clustering for Protein Representation Learning (CVPR 2024 · similarity 0.87)
- PSBench: a large-scale benchmark for estimating the accuracy of protein complex structural models (NeurIPS 2025 · similarity 0.87)
- AtomSurf: Surface Representation for Learning on Protein Structures (ICLR 2025 · similarity 0.87)
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.
