Research map / Probabilistic and scientific modeling / Molecules and proteins
Reading the Cell, Designing the Cure: Perturbation-Conditioned Molecular Diffusion for Function-Oriented Drug Design
ZIYU XU, zijian zhang, Liang Wang, Zhiyuan Liu, Qiang Liu, Shu Wu, Liang Wang
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 TFE-H heterogeneity-aware aggregation (ablation) | CURE · Tahoe-100M single-cell | Morgan Sim.: 0.4924 -> 0.6114 | Improved |
| add local fingerprint alignment (ablation) | CURE · Tahoe-100M single-cell | Morgan Sim.: 0.2674 -> 0.6114 | Improved |
| add global graph alignment (ablation) | CURE · Tahoe-100M single-cell | Morgan Sim.: 0.3219 -> 0.6114 | Improved |
| add TFE-A alignment on top of TFE-I (ablation) | CURE · L1000 bulk | Validity: 0.3000 -> 0.9350 | Improved |
| add TFE-I interaction on top of TFE-A (ablation) | CURE · L1000 bulk | Validity: 0.2400 -> 0.9350 | Improved |
| use TFE-I alone without alignment (ablation) | CURE · L1000 bulk | Morgan Sim.: 0.1824 -> 0.0886 | Hurt |
| use CURE instead of baselines on bulk in-distribution | CURE · L1000 bulk in-distribution | PRnet MSE: 2.5987 (Gx2Mol) -> 0.2328 | Improved |
| use TFE-H instead of pseudo-bulk averaging for single cell | CURE · Tahoe-100M single-cell | PRnet MSE: 4.1419 (Gx2Mol, pseudo-bulk) -> 0.4829 | Improved |
| use CURE on unseen cell lines | CURE · L1000 bulk unseen cells | Morgan Sim.: 0.6805 (GexMolGen) -> 0.9125 | Improved |
| use CURE for zero-shot gene inhibitor design | CURE · Gene inhibitor (EGFR) | Binding affinity (kcal/mol): -7.30 (GexMolGen) -> -9.11 | Improved |
| larger Top-K in drug screening with TFE embedding | CURE · Drug screener | Morgan Sim. vs Hit Rate: Morgan 0.9753 -> 0.9195; Hit Rate 0.6766 -> 0.9790 | Mixed |
| CURE versus Gx2Mol on QED | CURE · L1000 bulk in-distribution | QED: 0.6041 (Gx2Mol) -> 0.5665 | Hurt |
Closest papers by meaning
- Learning Cross-Domain Representations for Transferable Drug Perturbations on Single-Cell Transcriptional Responses (AAAI 2025 · similarity 0.87)
- Chain-of-Generation: Progressive Latent Diffusion for Text-Guided Molecular Design (NeurIPS 2026 · similarity 0.86)
- Reward-Guided Iterative Refinement in Diffusion Models at Test-Time with Applications to Protein and DNA Design (ICML 2025 · similarity 0.85)
- Controllable 3D Molecular Generation for Structure-Based Drug Design Through Bayesian Flow Networks and Gradient Integration (NeurIPS 2025 · similarity 0.85)
- Chem-PerturBridge: a harmonized compendium of small molecule perturbation transcriptomic effects (NeurIPS 2026 · similarity 0.85)
- TFG-Flow: Training-free Guidance in Multimodal Generative Flow (ICLR 2025 · similarity 0.85)
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.
