A Modified Skip-Gram Algorithm for Extracting Drug-Drug Interactions from AERS Reports.

Li Wang,Heming Bai,Qinghua Wang,Lei Jiang,Wenjie Pan,Yuanpeng Zhang,Wei Liu

doi:10.1155/2020/1747413

Abstract

Drug-drug interactions (DDIs) are one of the indispensable factors leading to adverse event reactions. Considering the unique structure of AERS (Food and Drug Administration Adverse Event Reporting System (FDA AERS)) reports, we changed the scope of the window value in the original skip-gram algorithm, then propose a language concept representation model and extract features of drug name and reaction information from large-scale AERS reports. The validation of our scheme was tested and verified by comparing with vectors originated from the cooccurrence matrix in tenfold cross-validation. In the verification of description enrichment of the DrugBank DDI database, accuracy was calculated for measurement. The average area under the receiver operating characteristic curve of logistic regression classifiers based on the proposed language model is 6% higher than that of the cooccurrence matrix. At the same time, the average accuracy in five severe adverse event classes is 88%. These results indicate that our language model can be useful for extracting drug and reaction features from large-scale AERS reports.

Highlights

Drug-drug interactions (DDIs) accounted for over 30% of all adverse drug events [1]
To assess the performance of the scheme based on modified skip-gram (MSG), we compared vectors generated from MSG with CM-Term frequencyinverse document frequency (TF-IDF)
We extended this idea to the enrichment of DDIs in DrugBank

Summary

Introduction

Drug-drug interactions (DDIs) accounted for over 30% of all adverse drug events [1]. More serious fact is that large quantity of DDIs manifested after a long period of exposure. Harpaz et al developed a taxonomy that characterized the associations and predicted several potential multi-item drug adverse effects [3]. They revealed that duplicate reports caused spurious associations. Cheng and Zhao integrated drug phenotypic, therapeutic, chemical, and genomic properties to predict DDIs [5]. These four types of drug-drug similarities were calculated as features of each drug-drug pair for prediction.

Methods

Results

Discussion

Conclusion