Fig. 3
From: HyperChIP: identification of hypervariable signals across ChIP-seq or ATAC-seq samples
Selecting a subset of genomic regions and using winsorization for parameter estimation. a For the H3K27ac ChIP-seq data set, bar plots show the distributions of top-ranked proximal/distal HVRs along the range of mean intensities. Proximal and distal regions have been separately divided into 10 equally sized groups based on the observed mean signal intensities. b Scatter plots of log10-scaled variances against the mean intensities of proximal/distal regions. Top-ranked HVRs and LVRs are separately highlighted. c Bar plots showing the distributions of top-ranked proximal/distal LVRs along the range of mean intensities. d d0 estimates resulting from different parameter estimation methods. The original method refers to the moment matching method employed by MAnorm2, which does not apply a subset selection and winsorization. The method labeled as “lower 10%” uses only the 10% of regions having the smallest mean intensities for parameter estimation