SLAM-Seq analysis

The sample preparation protocol for SLAM-Seq is described in Supplemental Experimental Procedures. SLAM-seq analysis was done according to a previous publication (Herzog et al., 2017) with minor changes. Briefly, we performed whole transcriptome sequencing instead of 3’ end sequencing. Reads are pre-processed with fastp (https://github.com/OpenGene/fastp) and then mapped by SlamDunk map, filter and snp (http://t-neumann.github.io/slamdunk/). T>C converted reads were separated by Alleyoop. The reads with T>C conversion were normalized by total mapped reads in each group. Due to the multimapper reconciliation strategy used in SlamDunk (multimappers are assigned towards 3’ end), there is a distort of the signal to the 3’ end of the gene. Metagene coverage matrix and profile were generated through deepTools (https://deeptools.readthedocs.io/en/develop/).