Similarity-based modeling in large-scale prediction of drug-drug interactions

Abstract

Drug-drug interactions (DDIs) are a major cause of adverse drug effects and a public health concern, as they increase hospital care expenses and reduce patients’ quality of life. DDI detection is, therefore, an important objective in patient safety, one whose pursuit affects drug development and pharmacovigilance. In this article, we describe a protocol applicable on a large scale to predict novel DDIs based on similarity of drug interaction candidates to drugs involved in established DDIs. the method integrates a reference standard database of known DDIs with drug similarity information extracted from different sources, such as 2D and 3D molecular structure, interaction profile, target and side-effect similarities. the method is interpretable in that it generates drug interaction candidates that are traceable to pharmacological or clinical effects. We describe a protocol with applications in patient safety and preclinical toxicity screening. the time frame to implement this protocol is 5–7 h, with additional time potentially necessary, depending on the complexity of the reference standard DDI database and the similarity measures implemented.

Revisiting Cross-Validation of Drug Similarity Based Classifiers Using Paired Data

“…The proposed models are typically benchmarked using cross-validation, in which the known drug-disease or drug-drug associations are split into training and test sets. Though these methods report areas under receiver operating characteristic (ROC) curves around 90% under cross-validation, their applicability in translational medicine and, thus, ability to reduce drug development costs has been controversial [2,4,5].…”

“…The proposed models are typically benchmarked using cross-validation, in which the known drug-disease or drug-drug associations are split into training and test sets. Though these methods report areas under receiver operating characteristic (ROC) curves around 90% under cross-validation, their applicability in translational medicine and, thus, ability to reduce drug development costs has been controversial [2,4,5].In light of previous works highlighting the perils of cross-validation using paired data [6,7], we recently investigated the effect of using drug-wise disjoint cross-validation in predicting drug-disease pairs, where none of the drugs in the training set appeared in the test set [8]. We showed that the prediction accuracy of the classifier drops dramatically under such cross-validation setting, suggesting that the existing approaches are prone to over-fitting due to the inherent relationships in the data.…”

Computational methods for inference and prediction of novel drug-drug interactions

“…Recently, they used the above three similarities added with drug-target similarity and adverse drug effects similarity to construct a new protocol [6] with the predicted effect correct in 36 out of 43 DDIs (84%). Not only added with two novel descriptors the new protocol also used principal component analysis (PCA) to show the results of combining the 5 different similarity measures with the STATISTICA software.…”

Reproducible Drug Repurposing: When Similarity Does Not Suffice

“…Nevertheless, in reality, the applicability of these methods for discovery of novel drug-disease associations has been limited due to "the reliance on data existing nearby in pharmacological space" as highlighted by Hodos et al 2 Moreover, Vilar and colleagues alert the community about the potential "upstream bias introduced with the information provided in the construction of the similarity measurement" in similarity-based predictors. 12 Yet, since many studies do not provide the data and code used to build the models for repurposing, it is often cumbersome to validate, reproduce and reuse the underlying methodology.…”