Data in RNA_seq_norm_iter0_hub was generated using Illumina HiSeq
* Raw 50-bp paired-end reads were trimmed using Trimmomatic (v.0.32) (Bolger et al. 2014; doi:10.1093/bioinformatics/btu170).
* Trimmed reads were mapped to human reference genome assembly GRCh38 using STAR aligner 
(v.2.4.1a) (Dobin et al. 2012: doi:10.1093/bioinformatics/bts635)
using the 2-pass procedure wherein novel splice junctions discovered on 
a first pass of alignments are compiled into a splice junction database used to inform a 
second pass of alignments as previously described (Engstrom et al. 2013; doi:10.1038/nmeth.2722). 
STAR aligner was run using default setting except with filtering for non-canonical novel intron junctions
and with "sjdbOverhang" parameter set to read length - 1.
* Mappings were normalized by RPKM as implemented in deepTools 
(Ramirez et al. 2014; doi: 10.1093/nar/gku365).