Preprocessing of reads was done automatically by seq2science v0.7.2 (https://doi.org/10.5281/zenodo.3921913) using the RNA-seq workflow. Paired-end reads were trimmed with fastp v0.20.1 (https://doi.org/10.1093/bioinformatics/bty560) with default options. Genome assembly GRCh38.p13 was downloaded with genomepy 0.12.0 (https://doi.org/10.21105/joss.00320). Reads were aligned with STAR v2.7.6a (https://dx.doi.org/10.1093%2Fbioinformatics%2Fbts635) with default options. Afterwards, duplicate reads were marked with Picard MarkDuplicates v2.23.8 (http://broadinstitute.github.io/picard). General alignment statistics were collected by samtools stats v1.14 (https://doi.org/10.1093/bioinformatics/btp352). Deeptools v3.5.0 (https://doi.org/10.1093/nar/gkw257) was used for the fingerprint, profile, correlation and dendrogram/heatmap plots, where the heatmap was made with options '--distanceBetweenBins 9000 --binSize 1000'. Sample sequencing strandedness was inferred using RSeQC v4.0.0 (https://doi.org/10.1093/bioinformatics/bts356) in order to improve quantification accuracy. Read counting and summarizing to gene-level was performed on filtered bam files using HTSeq-count v0.12.4 (https://doi.org/10.1093/bioinformatics/btu638). RNA-seq read duplication types were analyzed using dupRadar v1.20.0 (https://doi.org/10.1186/s12859-016-1276-2). Differential gene expression analysis was performed using DESeq2 v1.34 (https://dx.doi.org/10.1186%2Fs13059-014-0550-8). To adjust for multiple testing the (default) Benjamini-Hochberg procedure  was performed with an FDR cutoff of 0.1 (default is 0.1). Counts were log transformed using the (default) shrinkage estimator apeglm v1.16 (https://doi.org/10.1093/bioinformatics/bty895). TPM normalized gene counts were generated using genomepy based on longest transcript lengths. Quality control metrics were aggregated by MultiQC v1.11 (http://dx.doi.org/10.1093/bioinformatics/btw354).