Fig. 5: Association of the SE-GPS at increments of 0.3 with severe drug side effects in the Open Target and OnSIDES datasets.

The Open Target and OnSIDES datasets were restricted to drugs with a boxed warning or drugs withdrawn due to toxicity risk and the side effect phecodes matching the toxicity class. The association of increasing SE-GPSs with these severe drug side effects was investigated by binning the boxed warning dataset into 0.3 increments of the SE-GPS and comparing SE-GPS greater or equal to each increment with SE-GPS equal to zero. A logistic regression model was performed for each increment bin with drug side effect as the outcome variable and the SE-GPS bin as the predictor variable, adjusting for phecode categories as covariates. The statistical test was two-sided and ORs with 95% CIs are defined in the forest plot as circles and error bars, with filled circles indicating an OR with a significant P-value < 0.05 after correcting for multiple testing. Points are colored along a blue-to-red gradient, with blue representing lower OR values and red representing higher OR values. Panel a displays results for Open Targets (n = 69,290 independent drug– gene–phenotype combinations) and panel b displays results for OnSIDES (n = 30,652 independent drug– gene–phenotype combinations). The gray vertical line represents the null odds ratio (OR = 1). CI confidence interval, OR odds ratio, SE-GPS side-effect genetic priority score.