Fig. 2

(a) \(\:{C}_{t}\) values from the Allègre and Guo scRT-qPCR analyses for Nanog, Pou5f1 and two reference genes used in each study (Rpl30 and Rps17). The variations in the level of expression of the reference genes are of the same order as those of the genes of interest in the two datasets. (b) PCA graphs of the two main components for all datasets, considering the 8C, 16C, 32C, 64C and 90C stages for the 21 genes common to all datasets. Except for Goolam and Posfai’s dataset, a bifurcation event where an initially homogeneous population divides into two sub-populations as the developmental stage increases is visible, despite the reduced number of data and the minimal pre-processing procedure (see Methods). The rightmost panel shows, for Allègre’s dataset, that the two sub-populations are composed of two almost orthogonal sets of vectors, one defined by Epi lineage markers (Nanog, Fgf4, Pecam1, Sox2 and Bmp4), the other by PrE lineage markers (Fgfr2, Sox17, Pdgfra and Gata4). (c) Spearman correlation coefficients for Allègre’s and Guo’s datasets, using the pre-processing procedure described in the Methods section. The increase in the numbers of correlations and anticorrelations aligns with observations reported in the original studies (see text).