Fig. 2

Comparison of cDNA and dRNA sequencing results of potential poly(A) sites supported by more than 10 reads. (a) The proportion of potential pA sites supported by dRNA sequencing for the HCMV (purple, n = 181) and human (orange, n = 30,139) datasets. (b) Performance of the different filtering methods. The left side shows the positive predictive value of the internal-priming (IP - red) and template-switching (TS - blue) filters based on the dRNA sequencing results (positive predictive value ~ kept sites which are also detected in dRNA Seq). Potential human pA sites were filtered using SQANTI (yellow) and also based on whether or not they occurred in PolyA_DB (green). The right side of the panel shows the proportion of potential pA sites filtered out by the different filtering options not supported by dRNA sequencing (~ negative predictive value). (c) Barplot of the number of potential pA sites and regions with different adenine content in the HCMV (left) and human datasets (right). The features that the filtering algorithm characterized as TES are marked in blue, whereas putative artifacts are marked in red. (d) The positive predictive value of the different filtering methods is shown as a function of adenine content. The HCMV results are not detailed because the low number of TESs contained in the dataset cannot provide for a meaningful analysis