Fig. 3
From: Uncovering distinct protein-network topologies in heterogeneous cell populations
Unmixing observations of a mixed cell population by UNPBN in comparison to clustering approaches. a Mixtures of observations of EGF and NGF stimulated cells with different noise levels were generated as described. Observations were sampled at one-minute intervals for 10 minutes after stimulation. For each noise level and sampled time point, observations were unmixed using UNPBN, k-means clustering (with k=2) and hierarchical clustering (taking the final two clusters). The percentages of correctly allocated observations, averaged over all time points, are indicated by boxplots for the different methods as a function of the noise level (line within the box, the median; box, the 0.25 and 0.75 quartiles; whiskers, the largest and smallest data points which are still within the interval of 1.5 times the interquartile range from the box). b Comparison of the unmixing accuracy with noise level 0.7 along the different sampled time points, as achieved by UNPBN, post-processed UNPBN limited to two components, k-means (with k=2) and hierarchical clustering (taking the final two clusters)