GAIRD

Introduction

This project is an implementation of Graph reasoning method based on affinity identification and representation decoupling for predicting lncRNA-disease associations (GAIRD).

GAIRD designed homogeneous and heterogeneous distribution learning modules to combine information from different neighborhood scopes, and a representation decoupling strategy was established to distinguish the contributions of node attributes and network topology to the lncRNA-disease association prediction task.

Citation

If you find this repository useful, please consider citing our paper:

@inproceedings{
xuan2022prcs,
title={Graph reasoning method based on affinity identification and representation decoupling for predicting lncRNA-disease associations},
author={Shuai Wang, Hui Cui, Tiangang Zhang, Peiliang Wu, Toshiya Nakaguchi, Ping Xuan},
booktitle={Journal of Chemical Information and Modeling(under review)},
year={2023}
}

catalogs

/config: the initialization parameters for GAIRD.
/utils: tool used, e.g. dataset splitting, etc.
/data: dataset used in our study.
/model: code implementation of the GAIRD algorithm.
/output: output directory storing preprocessed features, segmented dataset, trained model, and prediction result.
main.py: scripts for model training and testing.
preprocessing.py: scripts for data preprocessing.

Environment

The codes of GAIRD are implemented and tested under the following development environment:

python == 3.6
networkx == 2.5
torch == 1.9.0
numpy == 1.19.2
scikit-learn == 1.0.2
matplotlib == 2.2.2

Dataset

disease_name.txt: disease names.
lncRNA_name.txt: lncRNA names.
disease_similarity.txt: disease similarity matrix computed from directed acyclic graphs(DAG) between diseases.
miRNA_similarity.txt miRNA similarity matrix obtained by similarity calculation for a set of diseases associated with two miRNAs.
lncRNA_disease_association.txt: lncRNA-disease associations extracted from the LncRNADisease database.
miRNA_disease_association.txt: miRNA-disease associations extracted from the HMDD database.
lncRNA_miRNA_interaction.txt: lncRNA-miRNA interactions extracted from the starbase v2.0 database.

How to Run the Code

Data Preprocessing: generating training set, test set, adjacency matrix, attribute matrix, shortest path distance matrix
```
python preprocessing.py
```
Train and test the model.
```
python main.py
```

The other details can be seen in the paper and the codes.