Ordering of rowVar in drawing Heatmap

Question

[dazhudou1122]

Dear Biostars community,

I am trying to draw a Heatmap with specific genes from RNA seq results. I want to order the genes by row variations so that the up-regulated genes are on top while the down-regulated genes are at the bottom (B6 is the control group, B6+WY is the treatment group). I used the codes below for Heatmap drawing but the sorting seems to fail as the orders are quite random. Up-regulated genes are mixed with down-regulated ones. Can anyone help me fix it? I would really appreciate it! Code:

library("genefilter")
library(pheatmap)

topVarGenes <- order(rowVars(assay(rld)), decreasing = TRUE)
mat  <- assay(rld)[ topVarGenes, ]
mat  <- mat - rowMeans(mat)

selectGene <- c("Txnrd1","Cat","Gsr","Gsta3","Gstm1","Gstp1","Gstt2","Mgst1","Nqo1","Prdx1","Sod1","Prdx6","Ftl1","Gdf15","Slc11a1","Slc25a37","Slc40a1","Slc48a1","Hmox1","Fos","Prkcg","Abcc1","Aox1","Dnajb13","Gclc","Gclm","Abhd3","Csf2ra","Pla2g7","Plb1","Pla2g2e","Tnf","Plcb4","Nfkbie","Nfkb2","Rsad2","Maff","Herpud1","Acta2","Hp","Cp","Sod3")

dataSub <- mat[rownames(mat) %in% selectGene, ]
anno <- as.data.frame(colData(rld))

png("redox.png", 1500, 3000, pointsize=25)
pheatmap(dataSub, annotation_col = anno, cluster_rows=FALSE, cluster_cols=FALSE, fontsize=24)
dev.off()

[ATpoint]

Use ComplexHeatmap and refer to its awesome documentation. https://jokergoo.github.io/ComplexHeatmap-reference/book/

[dazhudou1122]

Is there a way to do it using pHeatmap? That way I can keep my previous heat map drown using this software.

[Ram]

There probably is, but ComplexHeatmap has the modularity that no other piece of heatmap software does. It will make your life a lot easier.

[zx8754]

See: How to add images to a Biostars post and edit your post with images.

[dazhudou1122]

Thank you! I have updated the post. I copied the wrong link before.

[asalimih]

You mean the order of genes be the same as in "selectGene'? then just change

dataSub <- mat[rownames(mat) %in% selectGene, ]

to

dataSub <- mat[selectGene[selectGene %in% rownames(mat)], ]

[dazhudou1122]

Thanks for the reply. No I meant that I would like to plots the genes in the selectGene then sort them according their log2FoldChange so up-regulated genes are on top while the down-regulated genes are at the bottom.

[asalimih]

So first you must perform a differential expression analysis using a tool like limma or DESeq2,... then extract log2FCs from the result. after that you can sort 'selectGene' based on logFC numbers.

[dazhudou1122]

I did a DE analysis with DESeq2. Can you show me how to extract log2FCs and sort based on these numbers. If will be great if you can provide some sample codes. Thank you!

[Ram]

Did you look the resultsNames() of the DESeq2 result dataset? Check out the results() function to extract a result, and look at the column names of such a result dataset.

[Fabio Marroni]

Try something like dataSub <- dataSub[order(dataSub$L2FC),] (assuming L2FC is the log2fold change column).

As an alternative, you might just let the algorithm cluster them for you by using pheatmap(dataSub, annotation_col = anno, cluster_rows=TRUE, cluster_cols=FALSE, fontsize=24)