# DCGCN

**Repository Path**: liang_mao_lin/DCGCN

## Basic Information

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

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-09-25
- **Last Updated**: 2025-09-25

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# DCGCN
Here is the code for paper **DCGCN: Dual-channel Graph Convolutional Network-based Drug-target Interactions Prediction Method with 3D Molecular Structure**.

# System Requirements
The source code developed in Python 3.8 using PyTorch 2.1.1 The required python dependencies are given below. DCGCN is supported for any standard computer and operating system (Windows/macOS/Linux) with enough RAM to run. There is no additional non-standard hardware requirements.

numpy==1.24.4

scikit_learn==1.3.2

torch==2.1.1

tqdm==4.66.1

# Installation Guide
It normally takes about 10 minutes to install a new conda environment on a normal desktop computer. Run the following code under the conda environment to create the new virtual environment and install the required packages.
    
    $ conda create --name DCGCN python=3.8
    
    $ conda activate DCGCN

    $ pip install numpy==1.24.4
    
    $ pip install scikit_learn==1.3.2
    
    $ pip install torch==2.1.1
    
    $ pip install tqdm==4.66.1

# Datasets
We evaluated the performance of the method on two public datasets: DrugBank dataset and Luo's dataset.
For Drugbank dataset, we provided the split datasets for:

(1) ablation experiments in folder **./dataset/drugbank/result/**.

(2) five-fold cross-validation in folder **./dataset/drugbank/result/CV5/**.

(3) cold-start experiments in folder **./dataset/drugbank/result/cold_drug/** and **./dataset/drugbank/result/cold_protein/**.

For Luo's dataset, we provided the split datasets for five-fold cross-validation in folder **./dataset/dtinet/result/CV5/**.

Due to storage space restrictions on github, you can download our dataset by visiting the link: https://drive.google.com/drive/folders/1sy-6yH8VBC6s1LnR8RpEG3WWUg-SW_rP?usp=drive_link

Unzip the dataset folder and place it in the root directory of the project to achieve: ./dataset/

# Training and testing
You can use **main.py** to train the model with **DrugBank and Luo's dataset**. 

Line 4 can assign the GPU devices. 

Line 294 can assign the dataset and the train file manually.

The program can automatically save the best-performing model to the path **./models/**.

After training, you can run **evalute.py** for testing. You need to assign the task and the test set at line 268-273. In addition, you also have to assign the model for test at line 283. In folder **./models/**, we have provided some models that have been trained to help you reproduce the experimental results in the paper.