Research map / Probabilistic and scientific modeling / Molecules and proteins
Hierarchical Structure-Property Alignment for Data-Efficient Molecular Generation and Editing
Ziyu Fan, Zhijian Huang, Yahan Li, Xiaowen Hu, Siyuan Shen, Yunliang Wang, Zeyu Zhong, Shuhong Liu, Shuning Yang, Shangqian Wu, Min Wu, Lei Deng
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 clip-based data augmentation (ablation) | HSPAG · 15.0% data: property prediction | average AUC: 82.98 -> 86.36 | Improved |
| add clip-based data augmentation (ablation) | HSPAG · 15.0% data: conditional generation | RMSE: 0.419 -> 0.263 | Improved |
| add clip-based data augmentation (ablation) | HSPAG · 15.0% data: generation | uniqueness: 0.985 -> 0.864 | Hurt |
| add molecular-level-only alignment instead of hierarchical alignment (ablation) | HSPAG · 15.0% data: conditional generation | RMSE: 0.285 -> 0.263 | Improved |
| add any clip alignment (ablation) | HSPAG · 15.0% data: property prediction | average AUC: 82.44 -> 86.36 | Improved |
| add decoder to the model (HSPAG vs HSPAG-ND) (ablation) | HSPAG · 15.0% data: property prediction | average AUC: 85.12 -> 86.36 | Improved |
| add pre-training (ablation) | HSPAG · BACE | ROC-AUC (%): 80.2 -> 88.1 | Improved |
| add pre-training (ablation) | HSPAG · Tox21 | ROC-AUC (%): 78.5 -> 80.9 | Improved |
| add challenging set on top of random-only training (ablation) | HSPAG · complex molecules | RMSE: 0.5207 -> 0.3906 | Improved |
| add challenging set on top of example-only training (ablation) | HSPAG · complex molecules | availability: 0.6157 -> 0.6845 | Improved |
| use HSPAG instead of GCPN for low logP target | HSPAG · outlier region logP -2.5 to -2 | success rate (%): 85.5% -> 33.3% | Hurt |
| use HSPAG instead of LIMO | HSPAG · molecular distribution learning | validity: 1.000 -> .9483 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- LIMO — LIMO: Latent Inceptionism for Targeted Molecule Generation (ICML 2022) · better on 1 of 2 shared results
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
- Atomas: Hierarchical Adaptive Alignment on Molecule-Text for Unified Molecule Understanding and Generation (ICLR 2025 · similarity 0.87)
- TRIDENT: Tri-Modal Molecular Representation Learning with Taxonomic Annotations and Local Correspondence (NeurIPS 2025 · similarity 0.87)
- UniMatch: Universal Matching from Atom to Task for Few-Shot Drug Discovery (ICLR 2025 · similarity 0.86)
- pCoMole: Pareto-Constrained Molecule Editing with Discrete Flows (NeurIPS 2026 · similarity 0.86)
- Physical Consistency Bridges Heterogeneous Data in Molecular Multi-Task Learning (NeurIPS 2024 · similarity 0.86)
- Hierarchical Graph Tokenization for Molecule-Language Alignment (ICML 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.
