Research map / Probabilistic and scientific modeling / Molecules and proteins
Predicting a Protein's Stability under a Million Mutations
Jeffrey Ouyang-Zhang, Daniel Jesus Diaz, Adam Klivans, Philipp Kraehenbuehl
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 |
|---|---|---|---|
| learn epistasis via aggregated mutation representations instead of summing single-mutation predictions (ablation) | Mutate Everything · PTMul | Spearman rs: 0.50 -> 0.53 | Improved |
| learn epistasis rather than additive single predictions (ablation) | Mutate Everything · cDNA2 stabilizing double mutations | nDCG: 0.25 -> 0.43 | Improved |
| use pretrained AlphaFold backbone instead of ESM2 (ablation) | Mutate Everything · S669 | Spearman rs: 0.47 -> 0.56 | Improved |
| use AlphaFold pretrained weights instead of from scratch (ablation) | Mutate Everything · S669 | Spearman rs: 0.36 -> 0.56 | Improved |
| finetune the backbone instead of freezing (ablation) | Mutate Everything · S669 | Spearman rs: 0.48 -> 0.56 | Improved |
| predict residual added to sum of single ddG instead of direct prediction (ablation) | Mutate Everything · cDNA2 | nDCG: 0.40 -> 0.43 | Improved |
| predict multiplicative scaling instead of additive residual (ablation) | Mutate Everything · cDNA2 | nDCG: 0.43 -> 0.17 | Hurt |
| sum aggregation vs product aggregation of mutation features (ablation) | Mutate Everything · cDNA2 | nDCG: 0.42 -> 0.43 | No clear effect |
| whiten features before summing (ablation) | Mutate Everything · cDNA2 | nDCG: 0.43 -> 0.25 | Hurt |
| ensemble Mutate Everything with MSA and Tranception (ablation) | Mutate Everything · ProteinGym (all) | Spearman rs: 0.38 -> 0.49 | Improved |
| evaluate on reversed mutations (no reverse augmentation) | Mutate Everything · S669 reverse | Spearman rs: 0.56 -> 0.49 | Hurt |
Closest papers by meaning
- Predicting mutational effects on protein binding from folding energy (ICML 2025 · similarity 0.89)
- MutaPLM: Protein Language Modeling for Mutation Explanation and Engineering (NeurIPS 2024 · similarity 0.86)
- Enhancing Protein Mutation Effect Prediction through a Retrieval-Augmented Framework (NeurIPS 2024 · similarity 0.84)
- Zero-shot protein stability prediction by inverse folding models: a free energy interpretation (NeurIPS 2025 · similarity 0.84)
- Learning to Predict Mutational Effects of Protein-Protein Interactions by Microenvironment-aware Hierarchical Prompt Learning (ICML 2024 · similarity 0.84)
- Evolution-Inspired Loss Functions for Protein Representation Learning (ICML 2024 · similarity 0.84)
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.
