Supplementary Figure 1: Mapped causal interactions using the pairwise hierarchical model.
From: High-resolution genetic mapping of putative causal interactions between regions of open chromatin
(a), An example of a peak pair in a causal or pleiotropic relationship. The inset heatmaps show posterior probabilities for the interaction categories among the three peaks. The left heatmap shows the posterior probability of peak pairs in a causal relationship PPCjk, where the row indicates upstream peak j and the column indicates downstream peak k, suggesting the leftmost peak A is likely to be the upstream peak for the others (B and C). The upper triangle element of the right heatmap indicates the posterior probability of pleiotropy and the lower triangle element indicates the posterior probability of the linkage category. Our model predicts a pleiotropic relationship at B and C, driven by the variant in peak A. (b), Numbers of different types of directed acyclic graphs (DAGs) found in a high-confidence set of causally interacting peaks (see Section 4.6 in the Supplementary Note for details). (c), The MB21D2 locus with causal interactions predicted across 60 peaks; the highest peak in the middle was inferred to be the master regulatory peak with the putative causal variant rs2886870 inside the peak. (d), The numbers of caQTLs mapped by the standard hierarchical model (blue) and causal peak pairs mapped by PHM (red) detected in downsampled data with sample sizes ranging from 20 to 80.
