We performed all cis-eQTL mapping for neurotypical controls (BAs, aged > 17 years; Table 1) using tensorQTL (v.1.0.7), which leverages graphics processing units to substantially increase computational speed93. Initially, we filtered low expression as described previously12 using the GTEx Python script (that is, eqtl_prepare_expression.py) with modifications for isoform-level genomic features (that is, transcripts, exons and junctions). This script retained features with expression estimates greater than 0.1 TPM in at least 20% of samples and aligned read counts of six or more. Additionally, this script used Python functions defined by rnaseqnorm.py to normalize counts with TMM, a Python port of the edgeR function.
