Hi,
I want to have an overall visualisation for accessible regions of my interest similar as what plotHeatmap (deeptools) can provide. However, this has become tricky as for this data I just have access to peaks matrix and fragments files from cellranger but not the bam files. I already have tried RegionMatrix + RegionHeatmap available in signac but I have some doubts about the overall outlook. This is because I am comparing cell clusters with relatively large imbalanced sample sizes (500 cells versus 30000 cells) and RegionHeatmap normalizes the fragments counts by total number of cells and the huge heterogeneity within each cluster is not taken into account. So, I barely can see any signal for the cluster with 30000 cells with different sets of regions and I assume this is because of the normalization approach. Is there any other tool/method I can use for this purpose considering the input files I have? Thanks!
For heatmaps I often do pseudobulks, so summing the raw counts of cells per cluster, and then normalize with something like edgeR. This gives equal contribution of each cluster, avoiding exactly what you see, so domination of clusters due to size.
Thanks so much! For clarification: you follow this approach for atacseq data and by edgeR normalization, do you mean TMM normalization?
yes, exactly. atac, rna, chip...it's all just counts per region/gene in a matrix after all if you break it down.