LabeledIn: Cataloging Labeled Indications for Human Drugs

In 2014, researchers developed LabeledIn, a catalog of labeled indications for human drugs, by analyzing drug labels from DailyMed and manually annotating them to identify drug-disease treatment relationships. They focused on 250 highly accessed drugs, which corresponded to 500 unique drug labels after reducing redundancy. The final dataset included 7805 drug-disease treatment relationships with an inter-annotator agreement of 88.35%. The study emphasized the necessity of human expertise in the annotation process due to the limitations of automatic methods. LabeledIn provides detailed context for indications and a fine-grained drug representation, which is crucial for semantic interoperability and identifying drugs with specific indications related to dose forms or strengths. The dataset and guidelines are publicly available, and the study highlights the potential of LabeledIn to improve machine-learning methods for automatic drug indication extraction and enhance online health resources.