Research map / Probabilistic and scientific modeling / Molecules and proteins
SLAE: Strictly Local All-atom Environment for Protein Representation
Yilin Chen, Tianyu Lu, Cizhang Zhao, Hannah Wayment-Steele, Po-Ssu Huang
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 FAPE loss to pretraining objective (ablation) | SLAE autoencoder · CATH test, <128 res | all-atom RMSD (A): 3.89 -> 1.12 | Improved |
| add Rosetta energy regression to pretraining objective (ablation) | SLAE autoencoder · CATH test, <128 res | all-atom RMSD (A): 3.26 -> 1.12 | Improved |
| use all-atom input instead of backbone-only input (ablation) | SLAE autoencoder · CATH test, <128 res | all-atom RMSD (A): 4.61 -> 1.12 | Improved |
| add side-chain reconstruction to backbone-only model (ablation) | SLAE backbone-input ablation · CATH test, <128 res | all-atom RMSD (A): 5.23 -> 4.61 | Improved |
| use 8 A encoder cutoff instead of 4 A (ablation) | SLAE encoder · StructTokenBench functional site | average AUROC (%): 70.04 -> 75.20 | Improved |
| use 8 A encoder cutoff instead of 10 A (ablation) | SLAE encoder · CATH test, <128 res | all-atom RMSD (A): 2.10 -> 1.12 | Improved |
| use post-hoc kNN codebook instead of end-to-end LFQ/VQ quantization (ablation) | SLAE latent tokens · CATH test | sequence recovery accuracy (%): 75.2 -> 99.4 | Improved |
| enlarge post-hoc kNN codebook (ablation) | SLAE quantization · CATH test, <128 res | all-atom RMSD (A): 2.96 -> 1.60 | Improved |
| use decoder hidden representation instead of continuous latent | SLAE downstream features · StructTokenBench flexibility | average Spearman rho: 48.19 -> 57.05 | Improved |
| add mutation-site differencing (ablation) | SLAE-finetuned · Megascale stability test | RMSE (kcal/mol): 0.73 -> 0.68 | Improved |
| add interface-wise embedding (ablation) | SLAE-finetuned · PPB-Affinity | RMSE (kcal/mol): 2.01 -> 1.86 | Improved |
| use SLAE embeddings instead of PLM and structure-token embeddings | NMR chemical shift head · BMRB backbone nitrogen | RMSE (ppm): 2.12 -> 1.88 | Improved |
Closest papers by meaning
- Atom-level Protein Representation Learning Improves Protein Structure Prediction (NeurIPS 2026 · similarity 0.86)
- Pre-Training Protein Bi-level Representation Through Span Mask Strategy On 3D Protein Chains (ICML 2024 · similarity 0.86)
- Full-Atom Peptide Design with Geometric Latent Diffusion (NeurIPS 2024 · similarity 0.86)
- A Systematic Evaluation of Co-folding Model Representations for Small-Molecule Learning (NeurIPS 2026 · similarity 0.86)
- TRIDENT: Tri-Modal Molecular Representation Learning with Taxonomic Annotations and Local Correspondence (NeurIPS 2025 · similarity 0.86)
- AtomSurf: Surface Representation for Learning on Protein Structures (ICLR 2025 · 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.
