RNA Sequencing Data Analysis
STAR aligner was used (splice junction annotations based on GENCODE version 27; www.gencodegenes.org) to map paired-end raw reads to human genome build GRCh38. Based on GENCODE v27, mapped reads and gene annotations were enumerated using featureCounts. edgeR bioconductor package was used to compute log2 transformed counts per million mapped reads (log2CPM) and log2 transformed reads per kilobase per million mapped reads (log2RPKM). Removal of genes with the log2CPM inter-quartile range (IQR) of which across all samples was less than 0.5 preceded analysis of differentially expressed genes. Pairwise comparisons of differentially expressed genes between AR and non-AR was performed using edgeR. P-values were corrected for multiple testing using the Benjamini-Hochberg procedure (Benjamini, 2010) and false discovery rate (FDR). Corrected p-values were presented with a significance threshold of 0.05. Differentially expressed gene analysis was carried out using edgeR with the eosinophil percentage as a covariate. Default parameters were applied to run all software listed above.