To create a final peakset, we subsampled and merged BAM-files separately for schizophrenia and control samples. For each, a total of one billion paired-end reads were used, corresponding to 6.99 million read pairs sample per schizophrenia sample and 7.94 million read pairs per control sample. We called peaks separately on these two merged BAM files, and merging these two peaksets into a single consensus peakset. Finally, we removed peaks overlapping ENCODE blacklisted regions of low read mappability or repeated sequence (https://raw.githubusercontent.com/mills-lab/svelter/master/Support/GRCh38/Exclude.GRCh38.bed). For each peak, we assigned the closest gene and the genomic context of an ATAC-seq OCR using ChIPSeeker24; transcript database was built by GenomicFeatures25 upon ENSEMBL genes.
