Fig. 3: Association of the SE-GPS with drug side effects in the Open Target and OnSIDES datasets by drug side effect groupings.

a Forest plot showing ORs with 95% CI for the association between the presence of a SE-GPS > 0 (binarized as 1) and drug side effects, adjusted for 16 phecode categories using logistic regression. This was performed across the full Open Target dataset (n = 1,254,900 independent drug–gene–phenotype combinations) with the OR colored in red and stratified by the number of gene targets per drug (1, 2–5, 5+; blue), the number of side effects per drug (1–4, 5–10, 10+; purple) and the number of drugs per side effect (1–10,11–30, 30+; green). For each feature, unique genes (red) and unique phenotypes (blue) are recorded on the y-axis. b Replication analysis of (a) using the OnSIDES dataset (n = 1,158,368 independent drug– gene–phenotype combinations). c Forest plot showing ORs with 95% CI for the association between the presence of a SE-GPS > 0 (binarized as 1) and drug side effects, stratified by phenotype category in the Open Target dataset. d Replication analysis of (c) using the OnSIDES dataset. The sample size of each stratified group is detailed in the Source Data. The statistical test was two-sided and ORs with 95% CIs are defined in the forest plot as circles and error bars. Filled circles indicate an OR with a significant P-value. The red dashed line represents the null odds ratio (OR = 1). CI confidence interval, N number, OR odds ratio, SE side effect.