Fig. 3.

Putative template-switching artifacts differ from putative transcriptional end sites. (a) The nucleotide composition of the regions surrounding (±50 nt) putative TESs and putative template-switching artifacts in the HCMV dataset (above) and the human dataset (below). Common polyadenylation motifs are marked on the top of the panel. Zero denotes the location of potential pA sites. (b) Polyadenylation signals detected upstream of TESs (blue) and putative artifactual pA sites (red). Data for human PAS usage taken from reference [26] are shown in purple. (c) Density plot of the distance between the detected PASs and potential pA sites at positions characterized as TESs (blue) and at positions characterized as artifactual sites (red). (d) Heatmap showing the proportion of reads ending at a given nucleotide in the vicinity (±10 nt) of a potential pA site. The values of all high-confidence (supported by > 10 reads) potential pA sites are averaged. Darker colors mean that a higher proportion of alignments ended at a given position. The separate cDNA sequencing experiments from the HCMV dataset are shown separately. (e) Poly(A) tail length distributions measured by cDNA at TES (above) and at artifactual sites (below). The medians are shown as vertical lines. Apart from the median values which may be somewhat dislocated by to A-rich regions, it is important to note that long poly(A) tails (> 40 nucleotides) are just as prevalent in the genuine and in the artifactual groups