Variation of gene content and gene expression in terms of relative quantitative expression and tissue/organ specificity is a substantial factor affecting phenotypic diversity. In crops, particularly in cereals, the pan-transcriptome and pangenome concepts are being increasingly investigated after the reference genomes were made available. Characterizing the gene expression presence-absence variation (ePAV), tissue- and varietal-specificity of tetraploid durum wheat (Triticum turgidumssp. durum) enables to investigate the association between the genotypic and phenotypic variation at an unprecedented level of precision. The current study presents the transcriptome analysis for 13 elite varieties representatives of the worldwide cultivated germplasm. Gene expression variation is described in relation to a highquality reference genome sequence assembly of durum wheat cv. Svevo (c/o International Durum Wheat Genome Sequencing Consortium). cDNA libraries were produced from roots and leaves at the seedling stage and from developing grains. In order to study the gene expression pattern, these RNA-seq libraries were aligned to the durum wheat genome and the transcript abundance was calculated. Overall, 75.0, 70.5 and 74.5% of high-confidence Svevo genes were expressed in the grain, leaf and root, respectively. Principal Component Analysis (PCA) analysis showed a clear gene expression clustering lead by tissues and by varietal ancestry. Hierarchical clustering based on PC scores clearly differentiated up-and down-regulated gene clusters based on tissues and varieties. Variance expression analysis projected on the Svevo assembly allowed us to identify the chromosome regions that drove the major expression variation patterns. Interestingly, clustering the gene expression profiles and the cultivar’s expression profiles evidenced several gene expression patterns related to the ancestry relationship among cultivars, particularly for the grain. The functional annotation of these gene clusters is in progress. Towards a more complete assembly of a pan-transcriptome in durum, the cultivar-specific reads that could not be mapped on the Svevo genome (4-30% referred to the Svevo Illumina sequencing data) are being de novo assembled.