Moreza009/drug_approval_prediction

Paper | | Code

DrugReasoner Dataset

This dataset contains a curated collection of approved (Phase IV clinical trial) and unapproved (pre-clinical trial) small molecules from the ChEMBL database, annotated for drug approval status. It was designed for training and evaluating DrugReasoner, a reasoning-augmented LLM for drug approval prediction.

• Label: 1 for approved drugs, 0 for unapproved small molecules.
• Features: Each molecule is paired with its five most similar approved and unapproved molecules, based on a similarity metric.
• Source: All data were retrieved from ChEMBL database.
• Purpose: To support research in drug approval prediction, molecular reasoning, and interpretable AI in drug discovery.

Data verification and processing

A dataset of small molecules was obtained from the ChEMBL database (version 35) including approved molecules in Phase IV clinical trials and molecules in the preclinical phase. Child molecules were replaced by their parent compounds to avoid duplication. A subset of 2,255 approved and 2,255 unapproved molecules was created by random undersampling of the unapproved class. The simplified molecular input line entry system (SMILES) structure was retrieved, and strings were canonicalized using RDKit (version 2023.9.5). The dataset was partitioned into training, validation, and test subsets (8:1:1) using a stratified sampling strategy to maintain class distribution across all splits. An independent external dataset includes data presented in the ChemAP paper (Cho, C. et al., 2024). This dataset contained 20 approved and 8 unapproved drugs. Three approved drugs overlapping with the training, validation, or test sets were removed, and the remaining molecules (17 approved and 8 unapproved) were used for external evaluation.

Sample Usage

To use DrugReasoner, you must first request access to the base model Llama-3.1-8B-Instruct on Hugging Face by providing your contact information. Once access is granted, you can run DrugReasoner either through the command-line interface (CLI) or integrate it directly into your Python workflows.

Prerequisites

• Python 3.8 or higher
• CUDA-compatible GPU (recommended for training and inference)
• Git

Setup Instructions

1. Clone the repository

   git clone https://github.com/mohammad-gh009/DrugReasoner.git
   cd DrugReasoner
   

2. Create and activate virtual environment

   Windows:

   cd src
   python -m venv myenv
   myenv\Scripts\activate
   

   Mac/Linux:

   cd src
   python -m venv myenv
   source myenv/bin/activate
   

3. Install dependencies

   pip install -r requirements.txt
   

4. Login to your Huggingface account You can use this instruction on how to make an account and this on how to get the token

   huggingface-cli login --token YOUR_TOKEN_HERE
   

🚀 How to use

Note: GPU is required for inference. If unavailable, use our Kaggle Notebook.

CLI Inference

python inference.py \
    --smiles "CC(C)CC1=CC=C(C=C1)C(C)C(=O)O" "CC1=CC=C(C=C1)C(=O)O" \
    --output results.csv \
    --top-k 9 \
    --top-p 0.9 \
    --max-length 4096 \
    --temperature 1.0

Python API Usage

from inference import DrugReasoner

predictor = DrugReasoner()

results = predictor.predict_molecules(
    smiles_list=["CC(C)CC1=CC=C(C=C1)C(C)C(=O)O"],
    save_path="results.csv",
    print_results=True,
    top_k=9,
    top_p=0.9,
    max_length=4096,
    temperature=1.0
)

Citation

If you use this dataset in your research, please cite our paper:

@misc{ghaffarzadehesfahani2025drugreasonerinterpretabledrugapproval,
      title={DrugReasoner: Interpretable Drug Approval Prediction with a Reasoning-augmented Language Model}, 
      author={Mohammadreza Ghaffarzadeh-Esfahani and Ali Motahharynia and Nahid Yousefian and Navid Mazrouei and Jafar Ghaisari and Yousof Gheisari},
      year={2025},
      eprint={2508.18579},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2508.18579}, 
}

Datast Card Author

Mohammadreza Ghaffarzadeh-Esfahani

Dataset Card Contact

You can contact with the Corresponding author of the paper: Ali Motahharynia ([email protected])