Figure 5.

Exploratory cross-validation models and final model performance on an independent dataset. Elastic net models given all reliable digital biomarkers were generated for clinical outcomes. To explore model potential, 15 cross-validation models were trained on two thirds of the training cohort and evaluated on the remaining third (violin plots). Next, a final elastic net model was built in the entire training cohort. This model’s performance was evaluated on the data it was trained on (dashed line, N_patient = 56, N_trial = 102) and on an independent validation cohort (yellow stars, N_patient = 52, N_trial = 92).