Table 2.

Model performance for baseline features predicting treatment outcome.

Model	AUC (95% CI)	Accuracy %	Sensitivity	Specificity	R2
Clinical predictors	.659 (.629, .689)	.615	.623	.607	.057
Clinical predictors + a priori SNPs	.662 (.632, .692)	.628	.642	.613	.059
Clinical predictors + elastic net SNPs	.655 (.625, .685)	.609	.623	.594	.049

Clinical predictors model includes sociodemographic and pre-treatment symptom variables only. The model in the second row adds SNPs selected a priori based on work by¹⁰ to the clinical predictors model. The model in the third row adds the SNPs identified by the elastic net feature selection to the clinical predictors model. For threshold-dependent metrics (accuracy, sensitivity, specificity), a probability threshold of 0.5 was used for classification.