Extended Data Fig. 1: Grid-search for regularization parameter selection within repeated cross-validation framework.

Plots in each row display training and validation F₂ scores for the prediction of different outcomes with regularization parameter C varying from 0.001 to 100. Plots in each column refer to different features included in the model, that is, clinical (a), (f), (k); clinical and genetic (b), (g), (l); clinical and binarized genetic (c), (h), (m); genetic (d), (i), (n); and binarized genetic (e), (j), (o). The dot marks the model with the highest validation score. In each plot, data is presented as mean values of n = 300 independent scores derived from subsets of n = 179 (BioMe) and n = 1,816 (GPC) observations in validation and n = 358 (BioMe) and n = 3,632 (GPC) observations in training. Validation scores are presented as mean values ± SD.