ИСТИНА

Mammalian tissues are made up of multiple cell types with widely divergent gene expression profiles. Here, we analyze by RNA-seq the transcriptional profiles of 38 human primary cells of different lineages and from different body locations and compared them to the expression profiles of human tissues performed by the GTeX consortium and to Human Body Map (HBM) data generated by Illumina. We found that, with appropriate normalization, the transcriptional profiles of tissues segregate by tissue type regardless of sequencing protocol, while the transcriptional profiles of primary cells cluster according to cell morphology rather than by the tissue of origin. In particular, endothelial and epithelial cells form large separate clusters. To find biomarkers responsible for this separation, we used constrained and unrestricted linear models to reconstruct the transcriptional profile of a tissue as a linear combination of transcriptional profiles of individual primary cells and found the closest prototype of each tissue among the combination of primary cells. The assessment of splicing uncovers ∼280,000 splice junctions, of which about a quarter are novel, and confirms that the differences between primary cells and tissues also remain on the transcript level. Our results indicate a high degree of variation in the transcriptional profiles of primary cells and tissues and suggest that the transcriptome of a tissue is not immediately reduced to the transcriptomes of the constituent cell types.