Fig. 3

Impact of transcript mappability on the performance of DE analysis with a simulated dataset. We divided transcripts into three equal-sized groups of low, middle, and high values of transcript mappability and true transcript abundance. Intervals of transcript mappability were as follows: low, [0.0107, 0.257); middle, [0.257, 0.472); and high,[0.472, 1]. Intervals of mean true transcript abundance (CPM) for DE step evaluation were as follows: low, [0.539, 2.65); middle, [2.65, 9.68); and high, [9.68, 5.54 ×10⁴]. Intervals of true transcript abundance (CPM) for quantification step evaluation were as follows: low, [0.250, 1.10); middle, [1.1, 5.01); and high, [5.01, 6.91×10⁴]. (A) Relationship between the AUC score and transcript mappability faceted by mean true transcript abundance. (B) Relationship between F1 score and transcript mappability during the alignment step. (C) Relationship between Spearman’s rho of CPM value and transcript mappability faceted by true transcript abundance. Metrics were calculated for all RNA types. Hs, HISAT; St, StringTie; Ba, Ballgown, Ka; Kallisto, Sl; Sleuth, Sa; Salmon, De; DESeq2, Sr; STAR, Rs; RSEM, EB; EBSeq, Th; TopHat2, Cu; Cufflinks (also see Table 1)