Peer Review of Computational Drug Repositioning Based on Side-Effects Mined from Social Media
February 2016
TLDR The study improved and was accepted despite initial concerns about data clarity, methodology, and potential overfitting.
In 2016, a study on computational drug repositioning using side-effects data from social media was reviewed. Reviewers raised concerns about the clarity of data division into training and testing, the experimental setting and methodology, and the potential for overfitting a Support Vector Machine (SVM) classifier. They also questioned the validity of the study's findings, suggesting that the method might yield better results if it used side-effect data from the same resources as other studies. The reviewers criticized the study for not discussing the inverse correlation between side-effect frequency and the likelihood of two drugs sharing a protein target, and for basing its limitations mainly on an unpublished paper. Despite these concerns, the paper was accepted after significant improvements.
