Fig. 2

TSS-Seq peaks do not explain DoGs. TSS-Seq data from NIH3T3 cells were downloaded from DBTSS [8], and TSS-Seq peaks in the last 1kb of all genes were extracted using bedtools. Bar graphs a–c show the mean (and standard deviation) percentage of genes in each bin of TSS-Seq scores (sum of tag counts in the last 1kb) as calculated from 1000 sub-samples of expression-matched pan-stress DoG-associated and non-DoG-associated genes. Boxplots d–f show the cumulative distribution of TSS-Seq peak scores in log10 scale according to three inclusion criteria: a, d No filter: median TSS-Seq scores are 11 for pan-stress DoGs and 8 for non-DoG genes, p = 0.009. b, d Excluding 113 genes that harbor either a snoRNA or a TSS of another transcript within their last 1kb: median TSS-Seq score is 11 for pan-stress DoGs and 8 for non-DoG genes, p = 0.048. c, e Excluding additional 545 genes with zero TSS-Seq tag count in their last 1kb: median TSS-Seq score is 20 for pan-stress DoG- and 21 for non-DoG-associated genes, p = 0.37