RNA-seq QC and data processing

Raw RNA-Seq sequencing reads, with Phred (Q) ≥ 35, were trimmed (ribosomal sequence removal, quality threshold 20, minimum sequence length 35) using fastqmcf (v1.0), yielding a mean untrimmed read depth of ∼20 million reads/sample. Subsequent filtered reads were then mapped to the human (hg38) or mouse (mm10) reference genome using STAR (v1.9) (Dobin et al., 2013). Gene and transcript expression were determined by aligning merged RNA-Seq reads to RNA isoforms (Cupcake collapsed) from Iso-Seq datasets using Kallisto (v0.46.0) (Bray et al., 2016) with default parameters as input to SQANTI2 (https://github.com/Magdoll/SQANTI2). Using mouse RNA-Seq reads, a transcriptome assembly was generated using Stringtie (v2.1.4) (Pertea et al., 2015) with mouse reference GENCODE gtf (vM22), annotated and filtered with SQANTI2 (v7.4) using default parameters.