Fig. 11

Method validation with finger sweat (left column) and blood plasma (right column) data from Brunmair et al., 2021 [20] and Panitchpakdi et al., 2021 [41] respectively. On panels A to D, the standard deviations of constants of absorption and elimination of caffeine and diphenhydramine (\(k_a^\text {caf}\), \(k_e^\text {caf}\), \(k_a^\text {DPH}\), \(k_e^\text {DPH}\)) between the three modeled subnetworks are plotted. The number of points per method corresponds to the number of concentrations time series present in both data sets (i.e., 37 and 10 for sweat and plasma, respectively). A one-sided Wilcoxon signed-rank test was used to test for significant differences. Panels E and F show the estimated concentration time series of caffeine and DPH plotted from the three different subnetworks. The lines are named after the second metabolite with a known kinetic present in the subnetwork; however, they all refer to C of caffeine and DPH. The colors of curves and the area between them indicate the results from normalization with PKM\(_{\text {minimal}}\) or MIX\(_{\text {minimal}}\), respectively