The increasing availability of large multi-tissue data sets which contain gene expression measurements across different tissues and individuals provided unprecedented opportunities to investigate transcriptome variation across tissues and individuals, and may reveal interactions between genes and tissues. The corresponding data set is a three-dimensional array: genes, individuals and tissues (or recording times). We present here the so-called “Partial Triadic Analysis”,(PTA), a well suited statistical tool to get a clear representation of a spatial series of matrices, one for each tissue. PTA is an extension of PCA and allows one to find a structure common to every matrix and to study its stability across tissues. PTA consists in three steps: i) the interstructure step, where are compared and analyzed the relationships between the different datasets, ii) the compromise step, where all datasets are integrated into an optimum weighted average, the compromise (or consensus) table, and iii) the intrastructure step, where the single-transcriptome are compared to the compromise in order to analyze commonalities and discrepancies. PTA was applied to transcriptomic data from the ANR (Agence Nationale de la Recherche) funded BoSexDim project, consisting in 19 embryo transcriptomes recorded at D40 (40 days after fertilization) and structured by sex (Male / Female) and type (in vivo / in vitro). These transcriptomes were recorded for four tissues (brain, liver, gonad and placenta). PTA shows a compromise structured by sex (first axis) and type (second axis). The same set of genes contribute the most to the sex structuration whatever the tissue. However, the differentiation of in vivo vs in vitro embryos was not made by the same genes according to tissues. Some genes showed an inconsistent, even contradictory behaviour, with an overexpression in one tissue and an underexpression in another one. This example highlights the power of the Partial Triadic Analysis for depicting the variability of the transcriptome structure across various tissues. Acknowledgement: This research was funded by the ANR French organization (BoSexDim project). We are grateful to the Bosexdim consortium members for producing the biological material.