Research map / Probabilistic and scientific modeling / Molecules and proteins
Deep Molecular Representation Learning via Fusing Physical and Chemical Information
Shuwen Yang, Ziyao Li, Guojie Song, Lingsheng Cai
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 chemical expertise to PhysNet (ablation) | PhysNet · QM7 conformation learning | Distance MSE: 0.504$\pm$0.013 -> 0.492$\pm$0.027 | Improved |
| Add chemical expertise to PhysNet (ablation) | PhysNet · QM9 conformation learning | Distance MSE: 0.197 -> 0.255 | Hurt |
| Replace stand-alone ChemNet with fused PhysChem (ablation) | ChemNet · QM7 property prediction | MAE: 59.6$\pm$2.0 -> 59.6$\pm$2.3 | No clear effect |
| Replace stand-alone ChemNet with fused PhysChem (ablation) | ChemNet · SARS-CoV-2 property prediction | Multi-AUC: 0.757$\pm$0.012 -> 0.792$\pm$0.013 | Improved |
| Replace specialist cooperation with straightforward multi-task training (ablation) | PhysChem · QM9 property prediction | Multi-MAE: 1.096 -> 1.639 | Hurt |
| Use real conformations at test time (ablation) | ChemNet · QM9 property prediction | Multi-MAE: 1.096 -> 1.040 | Improved |
| Use RDKit-generated conformations (ablation) | ChemNet · Lipop property prediction | RMSE: 0.577$\pm$0.016 -> 0.532$\pm$0.011 | Improved |
| Add RMSD loss alongside Conn-3 loss (ablation) | PhysNet (s.a.) · QM9 conformation learning | Distance MSE: 0.197 -> 0.218 | Hurt |
| Add RMSD loss alongside Conn-3 loss (ablation) | PhysNet (s.a.) · QM9 conformation learning | RMSD: 0.709 -> 0.483 | Improved |
| Increase the loss weight on conformation learning (ablation) | PhysChem · QM9 property prediction | property prediction performance | Mixed |
| Increase the generalized-space dimension (ablation) | PhysChem · QM7 conformation learning and property prediction | performance | Mixed |
Closest papers by meaning
- Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 2024 · similarity 0.87)
- LagNet: Deep Lagrangian Mechanics for Plug-and-Play Molecular Representation Learning (AAAI 2023 · similarity 0.87)
- HAGO-Net: Hierarchical Geometric Message Passing for Molecular Representation Learning (AAAI 2024 · similarity 0.85)
- Navigating Chemical Space with Latent Flows (NeurIPS 2024 · similarity 0.84)
- Enhancing Molecular Representation Learning Through the Combination of 3D and 2D Graph Machine Learning (Student Abstract) (AAAI 2025 · similarity 0.84)
- MolTailor: Tailoring Chemical Molecular Representation to Specific Tasks via Text Prompts (AAAI 2024 · 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.
