Fig 1.
Flowchart of database collection methods.
Fig 2.
Flowchart of ML methods excluding comparison with Decision trees.
Fig 3.
A) Heatmap comparison of Drug-ADR associations between the SIDER database (clinical trials) and the YCS (real-world reports) datasets across MedDRA SOC categories. B) Venn Diagram representing all drug data proportions of each category in heatmap.
Fig 4.
A) Heatmap of model predictions. B) Heatmap of actual data. C) Heatmap of similarities and differences in model predictions and actual data. D) Pie chart representing distribution of predictions on all drug data for drug-ADR associations across MedDRA SOC categories.
Fig 5.
Performance metrics of the 21 SOC models for Decision Trees and Random Forests with significant differences indicated in ROC AUC and accuracy.
Fig 6.
Random Forest feature importance results of the top 10 targets for the Psychiatric Disorder model.
Fig 7.
Summary of SOC category associations for the top 10 targets based on feature importance, sourced from DisGeNET.
Fig 8.
Target Drug Relations linked to the top 10 targets based on feature importance.