Chromvar_clustering

[paulasoler]

Hello @timoast !

I am trying to do a downstream analysis with the result provided by RunChromVAR. If I understand correctly, this function retrieves the deviation scores, the Z-score for each cell for each annotated motif. Then, using the following functions I have tried to know how the different cell types are related; so firstly, I have averaged the counts from chromVar for the cell identity and then I plotted (using pheatmap) the result of avgexpr_mat$chromvar.

avgexpr_mat <- AverageExpression(
seurat,
assays = "chromvar",
return.seurat = F,
group.by = "ident",
slot = "data")

pheatmap(avgexpr_mat$chromvar, scale = "row")

As you could see from the heatmap, there is a clear biases towards two cell type where the majority of the signal is accumulated. So, can I do this type of analysis with the result of chromVar? Do I have to correct the data?

Thanks for all

[timoast]

Rather than looking at every motif, I'd suggest finding differentially accessible motifs between the cell types and then only plotting the top n differential motifs per cell type.