Distance trees of expression profiles.

<p>We constructed neighbor-joining trees based on the correlation between expression values (FPKM>1.0) between samples, with 1 minus Spearman's rho defining the distance. Colors denote library construction methods (poly-A: blue, DSN: red). We divided transcribed loci into (a) protein coding genes with RNA-Seq support, either annotated by EnsEMBL in dog or EnsEMBL in the human orthologous regions. Replicates cluster together, so do the library constructions methods poly-A and DSN, as well as related tissues, such as heart and muscle; (b) antisense transcripts, that overlap at least one exon of a protein coding gene, as defined in (a). With the exception of testis, poly-A and DSN separate the samples, with both the poly-A and DSN sub-trees maintaining closer relationships between the related tissues heart and muscle; (c) spliced intergenic loci, excluding sequences that have coding potential. Similar to protein coding genes, the poly-A and DSN group by tissue first, with the exception of kidney DSN; and (d) intergenic and uncharacterized single-exon transcript loci. In this set, DSN and poly-A are, similar to antisense loci, the most dominant factor when grouping samples.