Research map / Label-efficient and robust learning / Recognition and long-tail
Greater than the Sum of Its Parts: Building Substructure into Protein Encoding Models
Robert Calef, Arthur Liang, Manolis Kellis, Marinka Zitnik
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 |
|---|---|---|---|
| substructure-tuning (active+binding+conserved sites) | SaProt (650M) · EC prediction | Fmax: 0.778 -> 0.839 | Improved |
| substructure-tuning | ESM-C 300M · EC prediction | Fmax: 0.688 -> 0.761 | Improved |
| substructure-tuning | ESM2-650M · EC prediction | Fmax: 0.755 -> 0.745 | Hurt |
| substructure-tuning | ProSST-2048 · GO:MF prediction | Fmax: 0.522 -> 0.567 | Improved |
| substructure-tuning | ESM-C 600M · Subcellular localization | accuracy: 0.713 -> 0.635 | Hurt |
| substructure-tuning | ESM2-650M · Subcellular localization | accuracy: 0.710 -> 0.749 | Improved |
| substructure-tuning | ProSST-2048 · zero-shot DMS variant effect | Spearman's rho: 0.507 -> 0.356 | Hurt |
| substructure-tuning | ESM2-150M · thermostability | Spearman's rho: 0.627 -> 0.582 | Hurt |
| substructure-tuning | ProSST-2048 · Human PPI | AUROC: 0.925 -> 0.883 | Hurt |
| tune with secondary structure only (ablation) | ESM-C 300M · EC prediction | Fmax: 0.688 -> 0.618 | Hurt |
| tune with homologous superfamily only (ablation) | ESM-C 300M · EC prediction | Fmax: 0.688 -> 0.805 | Improved |
| tune with active sites only (ablation) | ESM-C 300M · GO:MF prediction | Fmax: 0.429 -> 0.518 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ESM-C 300M — ESMC: MLLM-Based Embedding Selection for Explainable Multiple Clustering (AAAI 2026) · better on 5 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.88)
- ProteinBench: A Holistic Evaluation of Protein Foundation Models (ICLR 2025 · similarity 0.87)
- Pre-Training Protein Bi-level Representation Through Span Mask Strategy On 3D Protein Chains (ICML 2024 · similarity 0.87)
- From Mechanistic Interpretability to Mechanistic Biology: Training, Evaluating, and Interpreting Sparse Autoencoders on Protein Language Models (ICML 2025 · similarity 0.87)
- Protein Structure Tokenization: Benchmarking and New Recipe (ICML 2025 · similarity 0.87)
- ProSST: Protein Language Modeling with Quantized Structure and Disentangled Attention (NeurIPS 2024 · 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.
