Fig. 4

Impact of mappability on the performance of DE analysis with an experimental dataset. We divided the transcripts into three equal-sized groups (low, middle, and high) according to the values of gene- and transcript- mappability and qRT–PCR measurement. Intervals of gene mappability were as follows: low, [0.46, 0.961); middle, [0.961, 0.991); and High, [0.991, 1). Genes with a mappability of ‘1’ were excluded from this grouping to avoid one group being occupied by one value. Intervals of qRT-PCR measurements (mean relative expression for internal control gene) were as follows: low, [0.000512, 0.0621]; middle, [0.0621, 0.353); and High, [0.353, 38.8]. Intervals of transcript mappability were as follows: low, [0.001, 0.255); middle, [0.255, 0.508); and High, [0.508, 1.00]. (A) Relationship between Spearman’s rho and gene mappability determined by qRT–PCR. Metrics were calculated for qRT–PCR validated 502 genes (see “Materials and methods for details of list of these genes). The Kallisto-Sleuth pipeline was excluded from this evaluation because it cannot output the gene-level fold-change value. (B) and (C) show the number of DEs compared with regular and mock comparisons, respectively. Metrics were calculated for all transcripts as defined by GENCODE. Hs-St-Ba; HISAT-StringTie-Ballgown, Ka-Sl; Kallisto-Sleuth, Sa-De; Salmon-DESeq2, Sr-Rs-EB; STAR-RSEM-EBSeq, Th-Cu; Tophat2-Cufflinks