# iPiDA-SWGCN
iPiDA-SWGCN: identification of piRNA-disease associations based on Supplementarily Weighted Graph Convolutional Network

# Dependencies
* python3.6 with pytorch 1.9.0 and torch-geometric 2.0.3
* Nvidia GPU with CUDA 11.1
* numpy 1.19.5
* sklearn 0.24.2

# Data description
* In this study, we construct piRNA-disease dataset based on MNDR v3.0 dataset(http://www.rnadisease.org/). After removing duplicate associations, 11981 experimentally verified piRNA-disease associations involving 10149 piRNAs and 19 diseases are collected. 
* piRNA_information.xlsx：list of piRNA names and piRNAs' corresponding number in the dataset 
* disease_information.xlsx: list of disease names and diseases' corresponding number
* adjPD.npy: piRNA-disease association matrix, a(i,j) represents the association between i-th piRNA and j-th disease
* piSim.npy: piRNA-piRNA similarity matrix
* DiseaSim.npy: disease-disease similarity matrix
* PD_ben_ind_label.npy: benchmark dataset and independent dataset division, where the label 1 and -20 represent positive and negtive samples in benchmark dataset, label -1 and -10 represent positive and negtive samples in independent dataset

# Run step
1. run `basic_ML.py` in weights_computation' folder to train several basic predictors to score the piRNA-disease associations. 
2. run `com_weights.py` in weights_computation' folder to compute the piRNA-disease association weights by intergrating several basic predictors .
3. run `train_PD_ind.py` to train the GCN model and obtain the final prediction scores for piRNA-disease associations
4. run `evaluate.py` to evaluate the model performance with AUC and AUPR

#run sample 
Please execute the following commands in turns

1. python ../iPiDA-SWGCN_code/weights_computation/basic_ML.py --file_path ../dataset/ \
      --piRSim piSim.npy --DisSim DiseaSim.npy --adjPD adjPD.npy \
      --PD_ben_ind_label PD_ben_ind_label_balance_col.npy --basic_predictor RF

2. python ../iPiDA-SWGCN_code/weights_computation/com_weights.py --file_path ../dataset/ \
      --piRSim piSim.npy --DisSim DiseaSim.npy --adjPD adjPD.npy \
      --PD_ben_ind_label PD_ben_ind_label_balance_col.npy 

3. python ../iPiDA-SWGCN_code/GCN/train_PD_ind.py --file_path ../dataset/ \
      --piRSim piSim.npy --DisSim DiseaSim.npy --adjPD adjPD.npy \
      --PD_ben_ind_label PD_ben_ind_label_balance_col.npy --Epoch 1000 --lr 0.001 --weight_decay 1.0

4. python ../iPiDA-SWGCN_code/weights_computation/com_weights.py --file_path ../dataset/ \
      --adjPD adjPD.npy --PD_ben_ind_label PD_ben_ind_label_balance_col.npy -- Y_pred predicted_score_SWGCN.npy