#GATLGEMF：Predicting ncRNA-protein interactions based on line graph attention networks

The untils, data and result directories contain model codes, data sets and generated results, respectively. The depended python packages are listed in requirements.txt and environments.yml. The package versions should be followed by users in their environments to achieve the supposed performance.

python 3.6.13

torch==1.10.1+cu113 torch-geometric==2.0.3 matplotlib==3.2.2 networkx==2.4 numpy==1.19.5 pandas==1.1.5 scipy==1.5.4 scikit-learn==0.24.2

Before using Main.py, if you want to use node feature, you need to use the below bash command to obtain node feature.

    python node_feature.py --dataset DATASET

#Train the network The program is in Python 3.6.13 using [Pytorch] backends. Use the below bash command to run NPI-LGAT.

    python Main.py --dataset DATASET

The parameter of DATASET could be RPI369, RPI2241, NPInter4158，RPI7317 and NPInter v2.0. Then, GATLGEMF will perform 5-fold cross validation on the specific dataset.

#Indenpendent test We demonstrated the model in jupyter notebook. The independent_test.ipynb is in the script folder. The test data and test model are in the data folder.

(If using this code , please cite our paper.) GATLGEMF：Predicting ncRNA-protein interactions based on line graph attention networks