Heatmap for RNA-Seq data

Question

[chetana]

I'm trying to generate a heatmap for almost 15,000 genes expression data across different samples. But they are grouped into clusters based on their similar expression across samples and I wanna represent the clusters vs samples rather than the genes vs samples (since they're like 15,000 genes) in the heatmap. Is there a way to do it? I know the melt function in R would help me plot the heat map in the way I want but I'm not sure how to implement it. I expect the heatmap to look something like this [https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4489295/figure/F2/]. Thanks in advance.

[Devon Ryan]

Why not use the tool that produced that plot? I mean the figure itself is from the paper describing that tool.

[chetana]

I'm sorry if I was not clear, I already have my data grouped into clusters in an another column in the text file. So my input data would be Genes, Clusters, Sample1, Sample2, Sample3,...so I want clusters vs samples on the heatmap. The paper doesn't mentioned about the data which has already been clustered.

[macmath]

Can you provide me details of how you grouped them into clusters, because I wish to group my genes into cluster and also extract the gene set for each differentially bound clusters.

[Chirag Parsania]

Provide your link. It's not there

[chetana]

Sorry about that, here is the link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4489295/figure/F2/

Answer 1

[Jake Warner]

Hi. You can achieve this by calculating the cluster core/ medoid and using those as input into heatmap2 or whatever your preferred heatmap function would be.

I had this function from Michael Dunrop bookmarked to calculate the cores:

clust.core = function(i, data, clusters) {
  ind = (clusters == i)
  colMeans(dat[ind,])
  }

clusters <-cutree(hr, h=1.5)
cores <- sapply(unique(clusters), clust.core, scaledata, clusters)

In the above, cutree generates clusters by cutting the gene dendrogram at 1.5. Replace clusters with whatever vector contains your cluster assignments.