Fig. 2

Amounts of polyA+ RNA in cells and effect of normalization. a Relative human polyA+ RNA amounts in samples were estimated by relating the human-specific sequence counts to the spike-in sequence counts. The amounts polyA+ RNA detected were EKC>LKC=HaCaT>SG. The EKC samples contained approximately twice as much polyA+ RNA as the SG samples. The samples contained equal concentrations of cellular RNA. b Comparison of gene expression between SG and EKCs of donor 10k by applying the gene-based normalization as implemented in SAMseq. Gene expression levels are shown as light gray dots (left panel), levels of spike-in RNAs as asterisks, and three control genes with distinct symbols. The spike-in RNAs appear twofold downregulated, although all samples contained equal amounts of spike-in RNAs. With the gene-based normalization, the Q-Q plot (right panel) shows approximately similar numbers of up- and down-regulated genes, shown as deviations from the diagonal. c The same comparison of gene expression between SG and EKCs by applying spike-in normalization as implemented in SAMstrt (left). Spike-in RNA counts follow the diagonal, but many more polyA+ RNAs appear upregulated than downregulated as shown by the upward shift of the gray cloud compared to (b). Q-Q plot (right) shows many more upregulated than downregulated genes, consistent with the increase of relative polyA+ RNA amount as shown in (a). d To experimentally validate which normalization yields a more correct analysis, qRT-PCR assays were performed on selected genes. Delta-Ct values indicate no change in spike-in RNA, approximately twofold upregulated RPL13A and RPLP0, and 8-fold upregulated GAPDH, consistent with the spike-in-normalized results. e Gene expression profiles for three sample types (SG, EKC, LKC) are similar between SAMstrt-normalized and RT-PCR results, whereas the SAMseq-normalized profiles show mild U shapes for RPL13A and RPLP0, different from the unchanged levels by RT-PCR and SAMstrt