Mapped reads were quantified on the level of genes using HTSeq version 0.6.1p1 64 and annotations from Gencode v19 56. We used the ‘union’ method of ‘htseq-count’ for unstranded libraries (-s no) and considered only uniquely mapping reads (-a 255; with 255 indicating uniquely mapped reads from the STAR aligner). Of note, STAR only outputs properly paired reads. Raw gene counts were scaled across individuals with scaling factors obtained with DESeq 65. The same approach was used to generate ‘probe-level’ counts using the final re-mapped set of gexarray probes. These were used to filter expressed probes in the CNA analysis. Finally, an alternative set of gene-level quantifications was generated for quality control purposes using RNA-SeQC v1.1.8 66. To match the original GTEx v6p quantifications as closely as possible, we ran RNA-SeQC with the -strictMode flag and used custom exon annotations generated and used by GTEx (gencode.v19.genes.v6p_model.patched_contigs.gtf.gz) to obtain gene RPKMs. The same RNA-SeQC quantification pipeline was applied to the two re-mapped GTEx tissues.
