# IDDGCN

**Repository Path**: Tomhappy/IDDGCN

## Basic Information

- **Project Name**: IDDGCN
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-01-08
- **Last Updated**: 2026-01-08

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Interpretable Dynamic Directed Graph Convolutional Network (IDDGCN) for Multi-Relational Prediction of Missense Mutation and Drug Response

'IDDGCN' explores the impact of missense mutations on drug responses by employing directed graphs to distinguish the relationships between sensitivity and resistance. It updates the weights of these relationships dynamically, using node features to reflect the specific interactions of different nodes. The model also integrates interpretability frameworks and proposes a method for establishing a reliable groundtruth to evaluate these interpretability models.

## Table of Contents

- [Framework](#framework)
- [System Requirements](#system-requirements)
- [Datasets](#datasets)
- [Run IDDGCN](#run-iddgcn)
- [Citation](#Citation)

## Framework
![Framework](image/IDDGCN.png)

## System Requirements
The source code was developed in Python 3.7 using TensorFlow 2.7.0. The required Python dependencies are listed below. IDDGCN is supported on any standard computer and operating system (Windows/macOS/Linux) with enough RAM to run. There are no additional non-standard hardware requirements.
```markdown
tensorflow>=2.7.0
scipy>=1.7.3
scikit-learn>= 0.22.1
numpy>= 1.21.6
pandas>=1.1.5
h5py>=3.8.0
openpyxl>=3.1.2
PubChemPy>=1.0.4
tqdm>=4.66.1
```

## Datasets
The `datasets` directory contains all the experimental data used in IDDGCN.
In the provided directories, we have full data with five-fold cross-validation experiment for prediction part and detailed data for explanation part:
- `datasets/prediction_datasets` contains all data for training IDDGCN and provides pre-trained weights in the `weights` subdirectory, which can be directly used for model evaluation.
- `datasets/explanation_datasets` contains all data needed to construct groundtruth and train the interpretability models, as well as the pre-constructed groundtruth and explanations generated by the interpretability models for evaluation.
## Run IDDGCN

1. **Model Training**: Run `IDDGCN.py` in the `prediction` folder to train the IDDGCN model. You can run this step to train the IDDGCN or directly proceed to step two to see the prediction evaluation results.
```markdown
cd prediction
python IDDGCN.py
```
2. **Model Evaluation**: Run `IDDGCN_eval.py` in the `prediction` folder to evaluate the prediction results.
```markdown
cd prediction
python IDDGCN_eval.py
```
3. **Ground Truth Construction**: You can use `ground_truth.py` in the `explanation` folder to construct groundtruth, or directly use the pre-constructed groundtruth `gt_all.csv` in the `explanation/data` folder.
```markdown
cd explanation
python ground_truth.py
```
4. **Run Explanation Models**: Directly run `Interpretability models` in the `explanation` folder to explain the prediction results. You can run this step to train these explanation models or directly proceed to step five to see the evaluation results.
```markdown
cd explanation
python GnnExplainer.py
python explaiNE.py
python IDDGCN_explain.py
```
5. **Explanation Model Evaluation**: Run `eval_test.py` in the `explanation` folder to evaluate the explanation models.  
```markdown
cd explanation
python eval_test.py
```
## Citation
If you use IDDGCN, please cite our paper: Qian Gao, Tao Xu, Xiaodi Li, Wanling Gao and Zhenyu Yue*. Interpretable Dynamic Directed Graph Convolutional Network for Multi-Relational Prediction of Missense Mutation and Drug Response, 2024, Submitted.