Extended Data Fig. 2. Effect of technical artifacts on eRNA capture.

a, A new strategy for evaluating strand specificity without the interference from promoter-upstream transcripts (PROMPTs)⁸¹. Red and blue colors indicate reads’ mapping direction; the highlighted (yellow) region indicates a previously validated⁸² PROMPT. Only the first exon in green was used for evaluation. b, Strand specificities of three stranded and unstranded RNA-seq libraries with our strategy. The p-value was estimated by a two-sided t test; c, Strand specificity for all libraries evaluated with our strategy. Values and error bars represent the mean and SD. $n=2$ (GRO-cap, CoPRO, csRNA-seq, PRO-seq, GRO-seq, mNET-seq), $n=3$ (STRIPE-seq), $n=4$ (CAGE and RAMPAGE), $n=8$ (BruUV-seq, total RNA-seq), $n=9$ (Bru-seq). d, Distribution of 3-mers at flush end sites⁸³ for RIP-seq and TGIRT-seq. The dashed red lines stand for the frequency of RT3-mers (sequence identical to the last three nts for the RT primer [for RIP-seq] or the 3′ adapter [for TGIRT-seq]) in the genome. e, Log odds ratios (LORs) of observed RT3-mer at flushing end sites versus in the genome (top) and internal priming rates (bottom) of assays when the internal priming could be detected from the sequencing data. f, The overlap between enhancers in the RppH library (Capped+Uncapped as “C+U”) that are also covered in the Capped library (C). The x-axis shows the minimum number of reads required for an enhancer locus to be considered as covered. g, Difference of log-transformed read counts between the capped (C) and RppH (C+U) libraries. The effect size was measured by Cohen’s d. In the box plot, the center dots, box limits, and whiskers denote the median, upper and lower quartiles, and 1.5× interquartile range, respectively. h, Pearson’s r of log-transformed reads from promoters of expressed transcripts (TPM > 5) was quantified using PRO-seq and POLR2A ChIP-exo. $n=\mathrm{4,747}$ (low), $n=\mathrm{9,058}$ (medium), and $n=\mathrm{2,470}$ (high).