Interacting and additive models are abundant and favor local genes. A. Density of Bonferroni Adjusted p-value distribution in randomly sampled intrachromosomal models. B. Density of non-additive interacting models where adj. p < 1x10-7, by minimum distance between regulatory elements. C. Distribution of model term retention across intra-TAD models. D & E. Euler plots of ATAC-seq peaks and genetic variants, dividing them by their participation in single-term, additive, and non-additive interacting models. ATAC-seq peaks and genetic variants involved in interacting models are highlighted, as they are investigated further in this paper.

ATAC-seq peaks that interact with genetic variants generally reside within the affected gene’s TAD. A. Schematic of a TAD loop, including gene (purple) and density of interacting model elements (red). Loop interior is in blue, exterior DNA is gray, and CTCF binding sites are in yellow. B. Location of interacting ATAC-seq peaks relative to TAD boundary location, merged across all genes. TAD interior denotes the TAD in which the dependent gene was found. C. Interacting ATAC-seq peaks by distance from associated gene TSS. Local area cutoffs of 100 kb and 500 kb flanking regions are marked.

TADs provide context for interactions and increase interaction search efficacy. A. Counts of intra-TAD ATAC-seq peaks involved in all non-additive interactive models, centered on the TSS of the gene affected by the genotype-ATAC interaction. Coordinates transformed to a standard scale. B. Example TAD, displaying interacting ATAC peak density and gene locations. Peak relevance generally decays relative to intra-TAD distance rather than linear chromosomal distance. C-F. A comparison between linear sequence-based and TAD-limited search methods for interacting ATAC-seq peaks. C and D compare percentage of significantly interacting ATAC-seq peaks at each gene-relative locus. E and F compare density of ATAC-seq peaks at each locus. TAD-based search shows a higher density of interactions and places limits on search distance due to testing only TAD-internal ATAC-seq peaks.

Motif analysis identifies differences in interacting CTCF binding motifs. A. A schematic of our motif analysis through MEMEsuite. FASTA files derived from interacting ATAC-seq peaks are used to identify enriched motifs, identify protein binding sequences, and locate the sequences within the ATAC- seq peaks. B-C. Binding sites found within significant motifs are less protected from genetic variation. SNP counts are shown at each locus in the CTCF binding sequence, comparing motifs within interacting ATAC-seq peaks versus all CTCF binding sites.

CTCF ChIP-seq analysis shows predictable strain-specific differences in binding intensity. A. Percentage of ChIP-seq peaks in surveyed strains. B. Variance (log10) in binding intensity fold enrichment for all ChIP-seq peaks. C. Percentage of significance in association between DO genotype at CTCF peaks and CTCF binding intensity on inbred ChIP-seq samples, in various subsets.