As isogenic iPSC lines showed a tendency to cluster in previous analyses (Figures 1C and S1E), we next selected the top 1,000 genes showing the largest variance in gene expression between all PSC lines (Table S3). Unsupervised hierarchical clustering using the 1,000 most variably expressed genes resulted in clustering of isogenic iPSC lines, supporting our previous observation that differences in gene expression were mostly donor dependent (Figure S4A). However, we also observed some variability between isogenic iPSC lines. GO analysis of these 1,000 genes indicated enrichment of categories representative of developmental pathways (Table S4). The oPOSSUM algorithm (Ho Sui et al., 2007) was used to identify regulatory motif over-representation across the most differentially expressed genes. This indicated hits in transcription factors related to the maintenance and differentiation of PSCs (Table S5). To analyze this further, we focused on an independent panel of genes associated with PSCs and their early differentiation, selected by the International Stem Cell Initiative (Adewumi et al., 2007). Unsupervised hierarchical clustering of iPSC lines according to the expression of these genes resulted in clustering according to the donor, confirming our previous findings that donor-dependent characteristics influence expression of genes related to pluripotency and differentiation (Figure S4B).
