Figure 2. Normalization of PAR-CLIP data shown for two representative EFs, Ctk2 (top) and Spt5 (bottom).

(A) Smoothed, raw RNA-binding strength as measured by the number of PAR-CLIP U-to-C transitions per U site for all mRNAs sorted by length and aligned at their RNA 5′-end (transcription start site, TSS). (B) Relative occupancy estimated by dividing the number of U-to-C transitions for each U site by the RNA-Seq signal at the corresponding genomic position for all mRNAs. A heat map showing the transcript-averaged RNA-Seq reads for all mRNAs scaled to the same length is shown below. (C) Relative occupancy estimated by dividing the number of U-to-C transitions for each U site by the Rpb1 PAR-CLIP signal at the corresponding genomic position for all mRNAs. A heat map showing the transcript-averaged Rpb1 PAR-CLIP reads for all mRNAs scaled to the same length is shown below. (D) Smoothed, raw and normalized PAR-CLIP signals as shown in A-C but averaged over all mRNAs. Before averaging RNA occupancy profiles were aligned at the RNA 5′-end and length-scaled such that the 5′-ends and pA sites coincided.

DOI: http://dx.doi.org/10.7554/eLife.25637.005

Figure 2—figure supplement 1. Normalization of PAR-CLIP data shown for two representative EFs, Ctk2 (top) and Spt5 (bottom), at mRNAs (left) versus CUTs (right).

Smoothed, raw and normalized PAR-CLIP signals as shown in Figure 2D but averaged over mRNAs (left) and CUTs (right). After normalization, average mRNA and CUT profiles were rescaled together, setting the maximum occupancy to one and the minimum occupancy to 0. This shows that after Pol II PAR-CLIP normalization Spt5 signals are equally high at mRNAs and CUTs, while no RNA normalization leads to less Spt5 signal at CUTs (due to less transcription of CUTs) and total RNA normalization leads to increased Spt5 levels at CUTs (due to decreased detection of unstable transcripts). Since Spt5 binds Pol II immediately downstream of initiation, differences in Spt5 RNA binding between mRNAs and CUTs would not be expected, arguing that Pol II normalization helps to correctly interpret the data.

DOI: http://dx.doi.org/10.7554/eLife.25637.006