Figure 3.

PCA and NMF analysis of the gene-expression data of mouse MSC, MAPC-1, MAPC-2 and ESC. (a) Percentage variation captured in each principal component (PC). PCs are ordered from 1 to 4 according to the percentage of the total variance they capture. (b) Samples plotted in the first two components' space. Distance of samples in the component space is indicative of similarity in expression profiles. (c) Consensus matrix from NMF. Model selection in NMF is based on a consensus matrix that contains the probability that a pair of samples is assigned to the same group. Probability values correspond to the colors in the key. M-1, MAPC-1; M-2, MAPC-2; (d) Metagene profiles from NMF plotted as logarithm of the probe set intensity.