Fig. 5
Model predictions of individuals in forensic scenarios. A) Illustration showing that the model is trained on a mean population behaviour and is now used to predict individual behaviours, a male (orange) and a female (pink). B-I) Model prediction of the four markers; B and F) blood alcohol concentration (BAC), C and G) urine alcohol concentration (UAC), D and H) blood ethyl glucuronide (EtG) concentration, and E and I) blood ethyl sulphate (EtS) concentration. The drinking-challenge for the male and female were consumption of an initial drink containing 0.85 g/kg ethanol composed of 13 v/v% wine over four 10 min block for a total of 60 min and a second drink containing 0.51 g/kg ethanol composed of 20 v/v% vodka consumed over 30 min. The solid line represents the best model solution from the model training, and the area is the model uncertainty. The data points are represented by the black circles, and the black bar indicates the period of alcohol consumption. J-Q) An in-silico drinking scenario for; J-M) the male and N-Q) the female, were evaluated. The alternative drinking-challenge was a single dose of 1.0 g/kg ethanol composed of 40 v/v% vodka for both the male and female. The model prediction is shown for the four different markers: J and N) BAC, K and O) UAC, L and P) EtG, and M and Q) EtS. The grey solid line represents the best model solution from the model training, and the grey area is the model uncertainty. The model prediction for the wine and vodka drinking-challenge (B-I) for the male is shown as reference for the male (orange J-M) and the female (pink, N-Q). The horizontal green lines indicate the regions where the model predicts differences between the two drinking-challenges for the respective individual (male and female).
