Predicting Frequencies of Drug Side Effects Using Graph Attention Networks with Multiple Features

A central issue in drug risk-benefit assessment is identifying frequencies of side effects. Frequencies were experimentally determined in randomized controlled clinical trials before, while it is time consuming and expensive. Recently, more and more computational models for predicting frequencies of drug side effects are put forward. In this work, we propose a novel method for predicting the frequencies of drug side effects, by using a Graph Attention Network to integrate different types of features, and integrating features embeddings from both drugs and side effects into a Multilayer Perceptron for prediction. The proposed method demonstrates performance through 10-fold cross-validation.