Research map / Probabilistic and scientific modeling / Molecules and proteins
Disentangled Wasserstein Autoencoder for T-Cell Receptor Engineering
Tianxiao Li, Hongyu Guo, Filippo Grazioli, Mark Gerstein, Martin Renqiang Min
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 |
|---|---|---|---|
| remove Wasserstein loss (ablation) | TCR-dWAE · TCR engineering ablation (VDJDB) | fraction positive valid: 0.23±0.02 -> 0.21±0.02 | Improved |
| remove Wasserstein loss (ablation) | TCR-dWAE · TCR engineering ablation (VDJDB) | MMD zf positive vs negative: 0.163±0.015 -> 0.040±0.026 | Improved |
| remove auxiliary classifier (ablation) | TCR-dWAE · TCR engineering ablation (VDJDB) | fraction positive valid: 0.23±0.02 -> 0.13±0.04 | Improved |
| remove auxiliary classifier (ablation) | TCR-dWAE · TCR engineering ablation (VDJDB) | MMD zf positive vs negative: 0.163±0.015 -> 0.087±0.067 | Improved |
| add decoder using only zs (ablation) | TCR-dWAE · TCR engineering ablation (VDJDB) | fraction positive valid: 0.23±0.02 -> 0.18±0.01 | No clear effect |
| add decoder using only zs (ablation) | TCR-dWAE · TCR engineering ablation (VDJDB) | MMD zf positive vs negative: 0.163±0.015 -> 0.238±0.029 | Mixed |
| use disentangled WAE instead of VAE | TCR-dWAE · VDJDB TCR engineering | fraction positive valid: 0.16±0.01 -> 0.23±0.02 | Improved |
| use TCR-dWAE instead of greedy / genetic / MCTS mutation baselines | TCR-dWAE · VDJDB TCR engineering | fraction positive valid: 0.02±0.0 -> 0.23±0.02 | Improved |
| replace zf with random embedding (ablation) | TCR-dWAE · VDJDB reconstruction | reconstruction accuracy: 0.9044 ± 0.0020 -> 0.7535 ± 0.0085 | Hurt |
| replace both embeddings with random (ablation) | TCR-dWAE · VDJDB reconstruction | reconstruction accuracy: 0.9044 ± 0.0020 -> 0.6016 ± 0.0030 | Hurt |
| use transformer decoder instead of LSTM decoder (ablation) | TCR-dWAE · VDJDB | fraction positive valid | No clear effect |
| model hypervariable region only (trimmed) (ablation) | TCR-dWAE · VDJDB | fraction positive valid | Hurt |
Closest papers by meaning
- Cross-Gate MLP with Protein Complex Invariant Embedding Is a One-Shot Antibody Designer (AAAI 2024 · similarity 0.85)
- Learning Disentangled Equivariant Representation for Explicitly Controllable 3D Molecule Generation (AAAI 2025 · similarity 0.84)
- SpecMER: Fast Protein Generation with K-mer Guided Speculative Decoding (NeurIPS 2025 · similarity 0.84)
- Full-Atom Peptide Design with Geometric Latent Diffusion (NeurIPS 2024 · similarity 0.84)
- ProSST: Protein Language Modeling with Quantized Structure and Disentangled Attention (NeurIPS 2024 · similarity 0.83)
- Learning to engineer protein flexibility (ICLR 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.
