Fig. 1From: Data integration by multi-tuning parameter elastic net regressionMean testing AUC as a function of the penalty ratio parameter κ for different simulation settings. The effect sizes and numbers of informative features are given in Table 1, Scenarios 1–5. Dots indicate the κ resulting in the maximum of mean testing AUC. Each analysis includes 200 samples per data set with 250 features per data type (N = 200 simulation replicates). When the number of informative features differs among the platforms (Scenarios 1–3), the multi-tuning parameter EN yields more predictive models comparing with the standard EN where κ=1. Differential penalization increased AUC the most when the effect sizes are smaller in the omic type with fewer informative features (Scenario 1)Back to article page