Clustering results used in the N-way HC-PLSR metamodelling with six clusters. A) Plot of the factors (Fac 1-Fac 3) from the global inverse metamodelling (=TOutput,A,Inverse). The observations are coloured according to cluster memberships. Cluster1=blue, cluster2=red, cluster3=yellow, cluster4=green, cluster5=magenta, cluster6=cyan. X is the 3-way state variable trajectory array, while Y is the parameters. The clustering was done on the factors explaining a significant amount of the variation in the state variable space, that is the 19 first factors. B) Plot of the predicted Y-scores (see Additional file 1, eq. S12c) from the global classical metamodelling, colour coded according to the cluster memberships of the observations found using TOutput,A,Inverse. The classification of the test set observations to be predicted in the classical metamodelling was based on , predicted from using second order polynomial OLS regression. This OLS prediction model was calibrated in the calibration step of the classical metamodelling, based on the factors from the inverse metamodelling (=, plotted in panel A) and calibration set (plotted here). C) Circadian clock state trajectories for the observations belonging to each cluster, coloured according to cluster memberships from the inverse N-way HC-PLSR. Cluster1 = blue, cluster2 = red, cluster3 = yellow, cluster4 = green, cluster5 = magenta, cluster6 = cyan. All state variables are given in nM units.
Tøndel et al. BMC Systems Biology 2012 6:88 doi:10.1186/1752-0509-6-88