TBPM-DDIE: Transformer Based Pretrained Method for predicting Drug-Drug Interactions Events
Zhenjie Shao, Ying Qian, Liang Dou · 2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC) · 2022
Millions of patients die from Drug-Drug Interactions (DDIs) each year and therefore DDIs have attracted widespread attention. A large number of deep learning-based methods have been used to testify whether there is an interaction between drugs in previous studies. However, few researchers pay attention to the specific interaction events between drugs. Recently, some new models for predicting DDIs events emerged, based on labeled drug pairs and leaving the unlabeled drug data not fully utilized. To make full use of unlabeled data to help predict the specific interaction events between drugs, we propose Transformer Based Pretrained Methods for improving the prediction of Drug-Drug Interactions Events (TBPM-DDIE) to extract a latent vector with drug structure and semantic information and then concatenate the vector and the diverse features of drugs as input to the DDIs events classifier. The TBPM-DDIE model can be divided into three parts, including the Transformer pretrained part, the Diverse features of drugs part, and the DDIs events classifier part. We have done experiments on the real-world dataset and compared it with the latest models. The results show that TBPM-DDIE can achieve state-of-the-art effects on predicting DDIs events.