Research map / Probabilistic and scientific modeling / Molecules and proteins
ProteinNPT: Improving Protein Property Prediction and Design with Non-Parametric Transformers
Pascal Notin, Ruben Weitzman, Debora Susan Marks, Yarin Gal
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 column-wise attention across labeled sequences (ablation) | ProteinNPT · 8 DMS assays, Contiguous CV | Spearman: 0.425 -> 0.501 | Improved |
| add column-wise attention across labeled sequences (ablation) | ProteinNPT · 8 DMS assays, Random CV | Spearman: 0.669 -> 0.684 | Improved |
| add column-wise attention across labeled sequences (ablation) | ProteinNPT · 8 DMS assays, Modulo CV | Spearman: 0.530 -> 0.531 | No clear effect |
| use labeled sequences at inference (100 vs 0) (ablation) | ProteinNPT · 8 DMS assays, Average | Spearman: 0.317 -> 0.569 | Improved |
| increase labeled sequences at inference beyond 1000 (ablation) | ProteinNPT · 8 DMS assays, Average | Spearman: 0.572 -> 0.572 | No clear effect |
| replace one-hot features with MSA Transformer embeddings (augmented) | ridge/baseline regressors · ProteinGym Contiguous CV | Spearman: 0.41 -> 0.47 | Improved |
| augment one-hot ridge regression with unsupervised fitness predictions | OHE baseline · ProteinGym Contiguous CV | Spearman: 0.08 -> 0.40 | Improved |
| replace Embed.-Aug. baseline with ProteinNPT | Embed. - Aug. (MSAT) · ProteinGym Random CV | Spearman: 0.57 -> 0.65 | Improved |
| replace Embed.-Aug. baseline with ProteinNPT | Embed. - Aug. (MSAT) · ProteinGym Contiguous CV | Spearman: 0.47 -> 0.48 | No clear effect |
| one-hot encodings used for cross-position extrapolation | OHE · ProteinGym Contiguous CV | Spearman: 0.54 (Random) -> 0.08 (Contiguous) | Hurt |
| add auxiliary labels (unsupervised fitness predictions) (ablation) | ProteinNPT · ProteinGym | Spearman | Improved |
Later papers that compared with this method
4 later accepted papers report ProteinNPT as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- RankFlow: Property-aware Transport for Protein Optimization (ICLR 2026 · better on 5 of 5 shared results)
- Metalic: Meta-Learning In-Context with Protein Language Models (ICLR 2025 · better on 2 of 2 shared results)
- Kermut: Composite kernel regression for protein variant effects (NeurIPS 2024 · better on 5 of 5 shared results)
- DePLM: Denoising Protein Language Models for Property Optimization (NeurIPS 2024 · better on 1 of 2 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
- A Variational Perspective on Generative Protein Fitness Optimization (ICML 2025 · similarity 0.88)
- ProSpero: Active Learning for Robust Protein Design Beyond Wild-Type Neighborhoods (NeurIPS 2025 · similarity 0.88)
- Multi-Scale Representation Learning for Protein Fitness Prediction (NeurIPS 2024 · similarity 0.88)
- CarbonNovo: Joint Design of Protein Structure and Sequence Using a Unified Energy-based Model (ICML 2024 · similarity 0.88)
- Tree Search-Based Evolutionary Bandits for Protein Sequence Optimization (AAAI 2024 · similarity 0.88)
- Learning to engineer protein flexibility (ICLR 2025 · similarity 0.88)
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.
