Figure 3

In vivo genomewide measurement of transcription elongation rates. (A) Schematic representation of the algorithm used to infer elongation rates, exemplified for the NDUFV2 gene. Left panel: 4sUDRB-seq data for 0/4/8 min after DRB release (similar to Figure 2B). Each signal was then corrected for the inferred background signal (right top panel), identifying for the 4 and 8 min samples the first position downstream to the TSS where the corrected signal is similar to that of the 0 time point; this location is designated by a vertical dashed line. To refine the boundary identification, we evaluated the background-corrected signals of the 4 and 8 min time points divided by the 0 time point signals and their derivatives. Refinement of the boundary position was then performed by monitoring the decreasing signal area in the vicinity of the rough boundary estimate, and determining the location between the closest peak and the point where the signal plateaus. (B) Linear fitting was performed on the averaged 4 and 8 min elongation boundaries as a function of time for the indicated genes. Elongation rates were defined by extracting the slope value of the linear fit (V). Confidence interval is indicated for each gene. (C) Distribution of measured elongation rates of all informative DRB-sensitive genes. (D) Box-whisker plot of log2 transformed H3K36me3 and H3K79me2 levels in genes that overlap with each modification in the 25% of genes with the highest calculated elongation rate and the 25% of genes with the lowest elongation rate. H3K79me2 is significantly higher in fast elongating genes (t-test, P = 1e-7). All data were adapted from ENCODE.