Fig. 4

piRNA expression pattern across human somatic, non-malignant tissues. a As an example of piRNA expression pattern heterogeneity across human somatic tissues, a similarity matrix was constructed using rank-normalized piRNA expression levels from breast (red), lung (blue) and stomach (turquoise) non-malignant human tissue. This data is derived from small RNA sequencing efforts from the The Cancer Genome Atlas (TCGA) consortium and processed through a computational pipeline to detect piRNA expression. Briefly, sequencing reads with a higher fraction of piRNAs (24–31 bp length) were mapped to the hg19/GRCh37 built. Aligned reads were subsequently quantified using specific piRNAs annotations, which account for the genomic locations of piRNA-encoding loci. Sample clustering was performed using Pearson correlation, and the correlation coefficient is color-coded from blue (low correlation) to red (high correlation). Noticeably, different degrees of overall correlation across samples from the same tissue origin are observed. Samples derived from breast tissue are mostly clustered together, denoting a high degree of similarity of piRNA expression patterns, while lung and stomach denotes subgroups. The molecular origin and physiological functions of these piRNA-defined subgroups is not completely understood, although it has been shown that these subgroups are linked to specific clinical features in cancer patients. b A subset of piRNAs is able to define a tissue-specific piRNA expression signature. In the figure, the top 10 over- and under-expressed piRNA on each tissue is compared against others. The expression is significantly different, assessed by comparative marker selection tool, with false discovery-rate (FDR) corrected p-value ≤ 0.05. Top panel: breast vs. lung + stomach; middle panel: lung vs. breast + stomach; bottom panel: stomach vs. breast + lung