Research map / Image generation / Autoregressive and latent generation
Cross-Chirality Generalization by Axial Vectors for Hetero-Chiral Protein-Peptide Interaction Design
Ziyi Yang, Zitong Tian, Yinjun Jia, Tianyi Zhang, Jiqing Zheng, Hao Wang, Yubu Su, Juncai He, Lei Liu, Yanyan Lan
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 axial feature injection (cross) to UniMoMo | UniMoMo (pep.) · LNR D-peptide design | right chirality % (minimized): 23.95 -> 99.81 | Improved |
| add axial feature injection (cross) to UniMoMo, L task | UniMoMo (pep.) · LNR L-peptide design | right chirality % (minimized): 76.98 -> 99.83 | Improved |
| add axial feature injection (triple) to UniMoMo | UniMoMo (pep.) · LNR D-peptide design | right chirality % (minimized): 23.95 -> 99.75 | Improved |
| add axial feature injection (commu.) to UniMoMo | UniMoMo (pep.) · LNR D-peptide design | right chirality % (minimized): 23.95 -> 99.88 | Improved |
| choice of axial vector type (cross vs triple vs commutator) (ablation) | PepMirror · LNR D-peptide design | average Vina score: -4.15, -4.20, -4.25 | No clear effect |
| add all three axial vector types concatenated (ablation) | PepMirror · LNR D-peptide design | IMP % (Vina): 72.04 -> 75.00 | Mixed |
| use pseudo-scalar triple_scalar instead of axial vector channels (ablation) | PepMirror (triple) · LNR D-peptide design | right chirality % (minimized): 99.75 -> 99.19 | Hurt |
| inject pseudo-scalar once instead of every layer (ablation) | PepMirror (triple_scalar) · LNR D-peptide design | right chirality % (minimized): 99.19 -> 96.98 | Hurt |
| add axial feature injection to the encoder (latent L-D discrepancy) (ablation) | UniMoMo encoder · LNR pocket latents | inversion-induced latent distance: 10^-2 scale median, more than four orders of magnitude larger than without AFI | Improved |
| switch from L to D design task | RFDiffusion · LNR Vina | average Vina score: -3.30 -> -1.77 | Hurt |
| switch from L to D design task | RFDiffusion · LNR Vina | IMP %: 44.09% -> 16.13% | Hurt |
| reflect test pockets across different planes instead of inversion (ablation) | PepMirror (cross) · LNR D-peptide design | IMP % (Vina): 63.44, 65.59, 69.89, 66.67 | No clear effect |
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.
- D-Flow — DFlow: A Generative Model Combining Denoising AutoEncoder and Normalizing Flow for High Fidelity Waveform Generation (ICML 2024) · better on 1 of 1 shared result
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
- UniMoMo: Unified Generative Modeling of 3D Molecules for De Novo Binder Design (ICML 2025 · similarity 0.86)
- Full-Atom Peptide Design with Geometric Latent Diffusion (NeurIPS 2024 · similarity 0.86)
- Full-Atom Peptide Design based on Multi-modal Flow Matching (ICML 2024 · similarity 0.85)
- PepTune: De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusion (ICML 2025 · similarity 0.84)
- E(3)-equivariant models cannot learn chirality: Field-based molecular generation (ICLR 2025 · similarity 0.84)
- Zero-Shot Cyclic Peptide Design via Composable Geometric Constraints (ICML 2025 · 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.
