Fig. 2: Analysis of the algorithm’s predictive components. | Translational Psychiatry

Fig. 2: Analysis of the algorithm’s predictive components.

From: Optimizing prediction of response to antidepressant medications using machine learning and integrated genetic, clinical, and demographic data

Fig. 2

A List describing the algorithm’s predictive genetic, clinical, and demographic components (within parentheses are the number of features that are mapped to each component, while the asterisks represent components whose mapped genetic features do not reside within the mentioned functional gene itself, but rather in an adjacent intergenic region), with marking per medication. B Pie chart describing the distribution of the selected components according to the two classes, following the scientific literature explorations. Only if associations of a component to depression and/or antidepressant were not found, we examined whether it was found to be associated with other psychiatric disorders or neurological phenotypes. C Top feature per medication’s final model (according to ROC curve variable importance or absolute coefficient size).

Back to article page