#22 · Epigenomics

Statistical frameworks for differential Hi-C and methylation region detection

The three items share a focus on statistical inference for epigenomic measurements rather than new mechanistic findings. For differential Hi-C, hicream applies pixel-level testing followed by clustering and post hoc analysis to recover differential regions of arbitrary shape. HiCPotts instead uses a Bayesian HMRF with zero-inflated Negative Binomial noise, via an ABC model, to identify significant Hi-C and micro-C interactions while modeling sources of bias. MultiDMPcaller addresses whole-genome bisulfite sequencing, detecting differentially methylated positions and regions through adaptive two-step P-value adjustment and multi-pairwise visualization. What is new across these tools is explicit treatment of noise, bias, and multiple testing in contact and methylation data, which bears on the reliability of condition comparisons. The dominant thread is methodological; the items do not establish biological conclusions.