Effective Knowledge Graph Embeddings based on Multidirectional Sementics Relations for Polypharmacy Side Effects Prediction

This is the code of paper Effective Knowledge Graph Embeddings based on Multidirectional Sementics Relations for Polypharmacy Side Effects Prediction. Junfeng Yao, Wen Sun, Zhongquan Jian, Qingqiang Wu, Xiaoli Wang.

Dependencies

• Python 3.6+
• Tensorflow 1.13.1+

Results

The results of MSTE on TWOSIDES and DrugBank are as follows.

TWOSIDES

	ROC-AUC	PR-AUC	AP@50
MSTE	97.44	96.73	98.86

DrugBank

	ROC-AUC	PR-AUC	AP@n
MSTE	99.59	99.48	99.37

Running the code

Usage

You can train the MSTE models on the two datasets by running its corresponding training script as follows:

for TWOSIDES dataset: python MSTE.py
for DrugBank dataset: python MSTE_DB.py

Citation

If you find this code useful, please consider citing our paper.

Acknowledgement

We refer to the code of TriVec. Thanks for their contributions.