Unsupervised clustering of super-enhancers from Chip Seq data generated from ROSE

Question

[Researcher]

Hi All, I have ranked my chip-seq peaks and calculated the probable super enhancers using ROSE for my samples from different subtypes of cancer. Now I want to perform unsupervised clustering based on enriched SE regions to see whether the samples from same cancer subtype cluster together or not. Can anybody please guide me how can I do this? Have someone done this before? Any help or recommendations will be highly appreciated. Thanks

[ATpoint]

Nothing special to be done. Follow the usual steps: Get a count matrix for the entire ChIP-seq dataset (all peaks), normalize e.g. with TMM from edgeR(please not RPKM), filter for the regions of interest, transform to Z-scale and then plug into hclust. I would start with ward.D2 as the agglomeration method and see if things look meaningful. I recommend ComplexHeatmap which has many convenience functions for clustering/visualization/plotting.

[Researcher]

Thanks ATpoint for your reply. But I am wondering about the sites that have different start and end loci but are overlapping. How should I deal with that?