Drug-drug interaction extraction via hybrid neural networks on biomedical literature

• Published in: Journal of biomedical informatics Vol. 106; p. 103432
• Main Authors: Wu, Hong, Xing, Yan, Ge, Weihong, Liu, Xiaoquan, Zou, Jianjun, Zhou, Changjiang, Liao, Jun
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.06.2020
• Subjects: Attention; Convolutional neural network; Drug safety; Drug-drug interaction extraction; Stacked bidirectional Gated Recurrent Unit; Drug-drug interaction extraction; Stacked bidirectional Gated Recurrent Unit; Drug safety; Convolutional neural network; Attention
• ISSN: 1532-0464; 1532-0480; 1532-0480
• DOI: 10.1016/j.jbi.2020.103432

SGRU-CNN model is established to extract drug-drug interactions (DDIs) from biomedical literature, which only takes word features and relative distance features as inputs, and accurately predicted relevant DDI types for targeting drug pair in the sentence. The model takes full of advantages of both the stacked bidirectional Gated Recurrent Unit (BiGRU) layers and the convolutional neural network (CNN) layer, which can be used to improve the effectiveness of classify pharmacological substances or extract DDIs from biomedical literature. [Display omitted] •Word embedding vectors were generated from the subset of MEDLINE.•A hybrid method for drug-drug interaction extraction from biomedical literature.•Different neural networks and pooling methods were applied to different features.•Two GRU networks were stacked to learn word feature and avoid overfitting. Adverse events caused by drug-drug interaction (DDI) not only pose a serious threat to health, but also increase additional medical care expenditure. However, despite the emergence of many excellent text mining-based DDI classification methods, achieving a balance between using simpler method and better model performance is still unsatisfactory. In this article, we present a deep learning method of stacked bidirectional Gated Recurrent Unit (GRU)- convolutional neural network (SGRU-CNN) model which apply stacked bidirectional GRU (BiGRU) network and convolutional neural network (CNN) on lexical information and entity position information respectively to conduct DDIs extraction task. Furthermore, SGRU-CNN model assigns the weights of each word feature to improve performance with one attentive pooling layer. On the condition that other values are not inferior to other algorithms, experimental results on the DDI Extraction 2013 corpus show that our model achieves a 1.54% improvement in recall value. And the proposed SGRU-CNN model reaches great performance (F1-score: 0.75) with the fewest features, indicating an excellent balance between avoiding redundant preprocessing task and higher accuracy in relation extraction on biomedical literature using our method.