Fig. 4

The success in identifying the correct number of distinct cell subpopoulations (i.e. components) in a mixture by UNBPN in comparison to clustering approaches. a A boxplot showing the ASW versus the tested number of components obtained by UNBPN analysis (here constrained in the postprocessing step to the imposed number of components) of a mixture of two subpopulations (EGF and NGF stimulated cells). The boxplot indicates the median (line within the box), the 0.25 and 0.75 quartiles (box), margined by the largest and smallest data points which are still within the interval of 1.5 times the interquartile range from the box (whiskers), and the outliers (dots) obtained from pooled values over all time points with noise level of 0.5. b and c, the same as in a but for ASW obtained following k-means clustering and hierarchical clustering, respectively. d, e and f, the same as the corresponding a, b and c, but for a mixture of 4 subpopulations: EGF-Mek ^wt, EGF-Mek ^mut, NGF-Mek ^wt and NGF-Mek ^mut (Additional file 1a-d). It should be noted that silhouette widths are incomparable between different clustering approaches. However, silhouette widths are comparable between different parameters of the same clustering approach and thereby indicate the identified number of distinct subpopoulations as the one providing the largest ASW