Sequencing quality was determined using FastQC [22] and 24 samples (11 cases and 13 controls) were excluded due to insufficient sequencing quality (e.g., strong overrepresentation of sequences, GC distribution). Raw reads were mapped to the human genome (hg38) using HISAT2 (v.2.1.0) [23]. Quantification was performed with the featureCounts function of the Rsubread package (v.2.0.1) [24], with hg38 annotation.
