🔮 qcMol: a large-scale dataset of 1.2 million molecules
with DFT-level quantum chemical annotations
for molecular representation learning

---

📘 Overview

qcMol is a large-scale and comprehensive dataset containing 1.2 million molecules collected from 95 curated sources.
It provides DFT-level quantum chemical annotations to advance molecular representation learning and molecular property prediction.

The dataset covers 247,448 unique molecular scaffolds, encompassing drug-like compounds, metabolites, and experimentally validated molecules, thus capturing a chemically diverse landscape suitable for AI-driven drug discovery.

Each molecule is annotated with detailed quantum-chemical descriptors derived from reliable B3LYP-D3/def2-SV(P)//GFN2-xTB calculations, combined with wavefunction analyses.
With multiple data formats and rich annotations, qcMol serves as both an ideal pretraining resource and a benchmark platform for deep learning models.

---

🌟 Dataset Highlights

• 🧩 Scale: 1.2 million molecules from 95 datasets
• 🔬 Quantum Calculations: B3LYP-D3/def2-SV(P)//GFN2-xTB
• 📊 Descriptors: NPA, ADCH, ELF, LI, LBO, DI, Mayer, and more
• 💊 Coverage: 247,448 molecular scaffolds across diverse drug-like and biological molecules
• ⚙️ Formats: Basic information / Global information / Atom-level information / Bond-level information
• 🧠 Applications: Pretraining, molecular property prediction, scaffold-level learning, and generative modeling

---

📕 Github Structure

qcMol/
├── README.md
├── env/                                             # Environment dependencies
│   ├── evaluation.md     
│   └── qc_calculation.md  
└── script/
    ├── evaluation/
    │   ├── descriptor_quality/                      # Experiment 1 - Features Quality
    │   │   ├── README.md             
    │   │   ├── gat_baseline/                        # Baseline setting
    │   │   ├── gat_conv6/
    │   │   ├── .....
    │   └── distribution_quality/                    # Experiment 2 - Distribution Quality
    │       ├── README.md
    │       ├── pretrain/                            # Pretrain with qcMol/QM9 data
    │       └── evaluation/                          # Evaluate the pretrained model
    └── qc_features_calculation/                     # The pipeline of quantum chemical calculation
        └── run_single/
            ├── ...
            ├── post_ana/
            └── result/   

---

📂 Access

• 🌐 Website: https://structpred.life.tsinghua.edu.cn/qcmol/
• 📄 Paper: https://www.biorxiv.org/content/10.1101/2025.09.07.674462v1

---

📚 Citation

If you use qcMol in your research, please cite:

@article{HaoyuWang2025qcMol,
  title     = {qcMol: A Large-Scale Dataset of 1.2 Million Molecules with High-Quality Quantum Chemical Annotations for Molecular Representation Learning},
  author    = {Haoyu Wang and et al.},
  journal   = {bioRxiv},
  year      = {2025},
  doi       = {10.1101/2025.09.07.674462},
  url       = {https://www.biorxiv.org/content/10.1101/2025.09.07.674462v1}
}