I was wondering if some one can help me with pseudo r code for my approach here based on this paper to plot a heatmap

Figure 3

enter image description here

Figure description The top 40 enriched genes percelltypeare shown in a heat map.Onlyhighlyexpressed genes with FPKM ⬎20 areincludedin this analysis. Fold enrichment iscalculatedasFPKM of one cell type divided by the average FPKM of all other cell types. The majority of these genes showed specific expression by only one cell type, with the exception that some are expressed during more than one maturation stage in the oligodendrocyte lineage.

Similar description in their methods: Differential expression was calculated as the FPKM of a given cell type divided by the average FPKM of all other cell types. Genes were ranked by their fold enrichment in each cell type, and top enriched genes for each cell type were identified

In my case I have 9 cell types(columns) and ~24000 genes(rows) from mouse and the values are TPMs.

I know to order celltype only based on one column like this

expr.order = expr[order(expr[,1],decreasing =T),]

Which clearly not the right way and would need some help with code in r

  • $\begingroup$ It’s unclear from your text what you actually want to do: there’s no question. For ordering the expression values based on one column you already found the solution. If that’s not what you mean you need to be more specific. $\endgroup$ Commented Sep 20, 2017 at 15:54
  • $\begingroup$ I do not know how to achieve this Genes were ranked by their fold enrichment in each cell type which I guess means rank individual column independently not sure how to do this $\endgroup$ Commented Sep 20, 2017 at 16:37
  • 1
    $\begingroup$ You can't rank individual columns independently (of the same data), but you can order them by several factors. $\endgroup$
    – llrs
    Commented Sep 21, 2017 at 11:23
  • 2
    $\begingroup$ Well in that case you first need to compute fold changes using one of the existing differential expression packages, e.g. DESeq2 or edgeR. $\endgroup$ Commented Sep 21, 2017 at 12:03
  • $\begingroup$ BTW, what else have you tried with R? Did you compute the fold change and found the top 40 as described in the image? $\endgroup$
    – llrs
    Commented Sep 24, 2017 at 20:52

1 Answer 1


The figure caption says "the top 40 genes per celltype" - so we can simply use the code you provided for the top 40 genes for a single celltype, repeat this for all celltypes, and then remove any duplicate genes. Here's an example:

> expr <- matrix(runif(9*24000), nrow=24000, ncol=9)
> top_genes_each_celltype <- sapply(1:ncol(expr), function(i) order(expr[,i],decreasing =T)[1:40])
> dim(top_genes_each_celltype)
[1] 40 9
> top_genes_all_celltypes <- unique(as.vector(top_genes_each_celltype))
> length(top_genes_all_celltypes)
[1] 358

Then you can build a biclustered heatmap with expr[top_genes_all_celltypes,], for example, using heatmap2.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.