Prediction of Drug-Drug Interactions by Using Profile Fingerprint Vectors and Protein Similarities


Abstract

Objectives: Drug-drug interaction (DDI) is a vital problem that threatens people''s health. However, the prediction of DDIs through in-vivo experiments is not only extremely costly but also difficult as many serious side effects are hard to detect in in-vivo and in-vitro settings. The aim of this study was to assess the effectiveness of similarity-based in-silico computational DDI prediction approaches and to provide a cost effective and scalable solution to predict potential DDIs.

Methods: In this study, widely known similarity-based computational DDI prediction methods were utilized to discover novel potential DDIs. More specifically, known interactions, drug targets, adverse effects, and protein similarities of drug pairs were used to construct drug fingerprints for the prediction of DDIs.

Results: Using the drug interaction profile, our approach achieved an area under the curve (AUC) of 0.975 in the prediction of a potential DDI. The drug adverse effect profile and protein profile similarity-based methods resulted in AUC values of 0.685 and 0.895, respectively, in the prediction of DDIs.

Conclusions: In this study, we developed a computational approach to the prediction of potential drug interactions. The performance of the similarity-based computational methods was comparatively evaluated using a comprehensive real-world DDI dataset. The evaluations showed that the drug interaction profile information is a better predictor of DDIs compared to drug adverse effects and protein similarities among DDI pairs.

Keywords: Computing Methodologies; Drug Interactions; Drug-Related Side Effects and Adverse Reactions; Knowledge Discovery; Pharmaceutical Databases.

Conflict of interest statement

Conflict of Interest: No potential conflict of interest relevant to this article was reported.

Figures

Figure 1
Figure 1. Prediction matrix M3 calculation.
Figure 2
Figure 2. Carrier target enzyme transporter (CTET) vector representation of drugs. Adapted from Ferdousi R, et al. J Biomed Inform 2017;70:54–64 [6].
Figure 3
Figure 3. Receiver operating characteristic (ROC) curve for interaction profile fingerprint test data. AUC: area under the curve.
Figure 4
Figure 4. Receiver operating characteristic (ROC) curve for adverse effect profile fingerprint test data. AUC: area under the curve.
Figure 5
Figure 5. Receiver operating characteristic (ROC) curve for protein test data. AUC: area under the curve.

Similar articles