Application of Biomedical Informatics Methods to Find Drug-Drug Interactions

Abstract

In order to improve patient wellbeing and advance clinical navigation, biomedical informatics technologies play a key role in identifying drug interactions (DDIs). Drug interaction (DDI) detection has emerged as a critical component of overall health security. Thus, applying text mining techniques to distinguish DDIs from biological writing has garnered significant attention. Nevertheless, this investigation is just getting started, and there is plenty of room to advance its presentation. We introduce a DDI extraction method based on syntactic convolutional neural network (SCNN). This approach suggests a new kind of word embedding termed syntax word embedding that takes advantage of the syntactic information present in a sentence. Knowing where each word is and how to highlight its grammatical forms allows us to expand its embedding. Because of this, the auto-encoder eventually figures out that the thick real worth vector is the standard word pack highlight (the inadequate 0-1 vector). The last step in removing DDIs from medical records is training the softmax classifier with a mix of conventional and embedding-based convolutional highlights. When compared to other state-of-the-art methods, SCNN performs better in terms of presentation, according to experimental results on the DDI Extraction 2015 corpus (F-score = 0.688).