Table 1.
Descriptive characteristics of the study population.
Fig 1.
Examples of simulated OGTT responses of study participant using the modeling scenarios.
Simulated glucose and insulin are depicted by blue and orange lines, respectively. The crosses denote the measured glucose and insulin concentrations.
Fig 2.
Explained variance (R2; panel A) and mean squared error (MSE; panel B) per glucose and insulin time-points following OGTT for the modeling scenarios.
Fig 3.
Individual parameter sets (n = 2968) in the parameter space of the Personalized eDES model.
The parameter space spanning the direction of the estimated model parameters k1, k5, k6 and k8 is shown after dimensionality reduction via principal component analysis. Dots represent the estimated parameter sets of individuals from the Maastricht Study and arrows represent the loadings.
Fig 4.
Relative feature importances in predicting the glucose concentration at the seven time points of the OGTT via the reference XGBoost models.
The 20 most important features are in decreasing order by relative feature importance at t = 0. The relative feature importances (x-axes) are calculated as the variance (MSE) reduction weighted by the proportion of samples reaching the node across all trees. Error bars represent the standard deviation across cross-validation folds.