Systematic drug repositioning based on clinical side-effects.

Area of Science:

• Pharmacology
• Biomedical Informatics

Background:

• Drug repositioning accelerates the exploration of new therapeutic uses for existing medications.
• Clinical side-effects (SEs) represent a human phenotypic profile associated with drug administration.

Purpose of the Study:

• To investigate the utility of clinical side-effects (SEs) as a predictive tool for drug repositioning.
• To establish a computational method for identifying novel disease indications based on drug-induced phenotypes.

Main Methods:

• Extracted 3,175 drug-disease relationships by integrating side-effect data from drug labels and PharmGKB.
• Developed Naïve Bayes classification models utilizing SEs as features to predict disease indications for 145 diseases.
• Validated the predictive performance of the models using Area Under the Curve (AUC) metrics.

Main Results:

• The models achieved an AUC greater than 0.8 in 92% of predictions for established drugs.
• The method demonstrated efficacy in predicting indications for clinical compounds, with 36% of models achieving an AUC above 0.7.
• Identified explicit repositioning hypotheses, such as using hypoglycemia-inducing drugs for diabetes treatment.

Conclusions:

• Clinical side-effects serve as a valuable 'clinical phenotypic assay' for rational drug repositioning.
• Attention to SEs during clinical trials can uncover significant repositioning potential beyond safety assessments.
• This strategy offers a systematic approach to discover new therapeutic applications for drugs and clinical candidates.