Predicting Drug Side Effects Using Data Analytics and the Integration of Multiple Data Sources

Abstract

The development of automated approaches employing computational methods using data from publicly available drugs datasets for the prediction of drug side effects has been proposed. This paper presents the use of a hybrid machine learning approach to construct side effect classifiers using an appropriate set of data features. The presented approach utilizes the perspective of data analytics to investigate the effect of drug distribution in the feature space, categorize side effects into several intervals, adopt suitable strategies for each interval, and construct data models accordingly. To verify the applicability of the presented method in side effect prediction, a series of experiments were conducted. The results showed that this approach was able to take into account the characteristics of different types of side effects, thereby achieve better predictive performance. Moreover, different feature selection schemes were coupled with the modeling methods to examine the corresponding effects. In addition, analyses were performed to investigate the task difficulty in terms of data distance and similarity. Examples of visualized networks of associations between drugs and side effects are also discussed to further evaluate the results.