Research map / Probabilistic and scientific modeling / Molecules and proteins
CAPSUL: A Comprehensive Human Protein Benchmark for Subcellular Localization
Yicheng Hu, Xinyu Lin, Shulin Li, Wenjie Wang, Fengbin ZHU, Fuli Feng
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 |
|---|---|---|---|
| pre-train ESM-C instead of random initialization (ablation) | ESM-C 600M · CAPSUL test set | micro-avg F1: 0.338 -> 0.495 | Improved |
| use ESM-C instead of ESM-2 650M | sequence-based language model · CAPSUL test set | micro-avg F1: 0.375 -> 0.495 | Improved |
| finetune the pre-trained ESM-C (ablation) | ESM-C 600M · CAPSUL test set | micro-avg F1: 0.495 -> 0.492 | No clear effect |
| use true C-alpha coordinates instead of random ones (ablation) | CDConv · CAPSUL test set | micro-avg F1: 0.329 -> 0.452 | Improved |
| use true C-alpha coordinates instead of random ones (ablation) | GearNet-Edge · CAPSUL test set | micro-avg F1: 0.348 -> 0.417 | Improved |
| add contrastive loss (ablation) | CDConv · CAPSUL test set | macro-avg F1: 0.226 -> 0.234 | Mixed |
| add early fusion with ESM-C | CDConv · CAPSUL test set | micro-avg F1: 0.452 -> 0.470 | Improved |
| add late fusion with ESM-C to CDConv versus ESM-C alone | ESM-C+CDConv · CAPSUL test set | micro-avg F1: 0.495 -> 0.476 | Hurt |
| add class reweighting (ablation) | CDConv · CAPSUL test set | macro-avg F1: 0.226 -> 0.297 | Mixed |
| add class reweighting (ablation) | GearNet-Edge · CAPSUL test set | micro-avg F1: 0.417 -> 0.453 | Improved |
| add class reweighting (ablation) | ESM-C 600M · CAPSUL test set | micro-avg F1: 0.495 -> 0.429 | Hurt |
| use AlphaFold2 instead of Boltz-2 predicted structures | CDConv · CAPSUL subset (1,223 proteins) | micro-avg F1: 0.477 -> 0.527 | Improved |
Closest papers by meaning
- Protein Structure Tokenization: Benchmarking and New Recipe (ICML 2025 · similarity 0.84)
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.84)
- PPDiff: Diffusing in Hybrid Sequence-Structure Space for Protein-Protein Complex Design (ICML 2025 · similarity 0.84)
- PSBench: a large-scale benchmark for estimating the accuracy of protein complex structural models (NeurIPS 2025 · similarity 0.83)
- Towards precision protein-ligand affinity prediction benchmark: A Complete and Modification-Aware DAVIS Dataset (NeurIPS 2025 · similarity 0.83)
- CryoDomain: Sequence-free Protein Domain Identification from Low-resolution Cryo-EM Density Maps (AAAI 2025 · similarity 0.83)
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.
