Improved Quality Metrics for Association and Reproducibility in Chromatin Accessibility Data Using Mutual Information

Correlation statistics are commonly used in the field of (epi)genomics and often applied with no regard to the underlying assumptions of the statistic(s), such as homoscedasticity, normality, and independence of values. This is true when analyzing values between replicated sequencing experiments that probe chromatin accessibility, i.e., assays for transposase-accessible chromatin via sequencing (ATAC-seq). To study the behavior of correlation measurements between ATAC-seq experiments, we generated sequencing data from three experiments using the A549, human lung cell line, totaling eight samples. Of these eight, three biological replicate libraries were prepared from freshly harvested cells using the Active Motif ATAC-seq kit, while the remaining five libraries were prepared using cryopreserved cells. Paired end sequencing was conducted on each of the eight ATAC-seq libraires, generating eight pairs of fastq.gz files. We developed a computational simulation with these data to explore the behavior of correlation metrics and to compare their accuracy under set conditions of reproducibility. These raw fastq.gz files are submitted here to NCBI's equence read archive.