Fig. 4

Probing the detection limit with 10 to single cells. a Representative denaturing PAGE (6%) of the PEG/NaCl-purified, amplified cDNA libraries from single hPSCs or sorted endothelial progeny. Differentiation, sorting, and STA were performed as in Fig. 3a. For single cells, only cDNAs successfully amplified after 21 cycles of preamplification (asterisks) were size-selected (rectangular box) for library construction. b The supervised heat map of total RNA expression of all 12 PSC and END samples. Genes (summed count across all 12 samples >20) from all 12 samples were used for differential-expression analysis with DESeq2. The rlog-transformed counts with lowest p values (1000) were ranked by log2 fold changes and served as input for heatmap3 without rearranging column and row dendrograms. c Scatter plots of rld_miRNA from individual samples of hESCs against the averaged rld_miRNA of six endothelial samples. Only miRNAs with summed counts >20 across 12 PSC and END samples were included for DESeq2 analysis. Colored dots represent miRNAs enriched in hPSCs (blue) or endothelial cells (red) as defined in Fig. 3d. The blue dots were transformed into squares when the values of (rld – average rld)/log10 (10 + average rld) were less than or equal to 0.4. The ratio of blue squares over total blue (square + dots) is indicated in the upper left corner. Exemplary aberrant expressers in low-input samples are highlighted with black borders. d rld of aberrantly detected miRNAs (highlighted dots in c) in individual low-input (arrows) vs those in the other samples of hESCs (P) and endothelial cells (E). e Scatter plots of rld_{protein-coding} from individual samples of hESCs (upper) and endothelial cells (lower) against the averaged rld_{protein-coding} of the six 293T samples. Colored dots represent protein-coding genes differentially expressed (p < 0.01) in hESCs (blue), endothelial cells (red), and 293T cells (black) by DESeq2 analysis. Protein-coding genes with summed counts >20 across the 12 samples (the 6 hESC and the 6 293T samples, or the 6 endothelial and the 6 293T samples) were included for differential-expression and rlog-transformation analyses. The blue and red dots were transformed into squares when the count of a particular gene was 0. The ratios of squares over (squares + dots) are indicated in the upper left corner. f Unsupervised PCA based on the mature miRNA counts of all 18 samples. miRNAs (mature miRNA by accession number, miRBase v21) with summed counts >500 across the 18 samples were rlog-transformed with DESeq2 (blind = TRUE). The transformed rld served as the input for PCA with default parameters. g Visualization of novel (panels 1, 2) and mitochondrial (panels 3–5) transcripts in the UCSC Genome Browser. Each curve represents reads per million (RPM)-normalized wiggle output of the libraries against the GRCh38 genome assembly. h Validation of the transcripts by denaturing PAGE (12%) for the novel (upper) and mitochondrial (lower) transcripts. One-hundredth of the preamplified cDNAs in Figs. 1g and 3f was used for 30 cycles of PCR