Most variable gene from the differentially expressed genes

Question

[1769mkc]

I m getting deseq2 output of differentially expressed genes with p ,0.05 .So from that list of differentially expressed i want to take out the most variable genes getting expressed .Is there any way to do that .

Any suggestion or help would be highly appreciated.

Answer 1

[Satyajeet Khare]

I think, by "most variable" you mean in terms of "differences in expression between the two samples". You need to have fold change values for that. Generally, genes with p<0.05 and log2FC > 1 (more than two fold) are selected.

[1769mkc]

well yes i did that too, but when i plot heatmap i dont see much differences,and lets say i have differentially expressed gene list of 1000 ,i dont think that would be wise to take into downstream analysis .I want to narrow it down..make it more stringent

[Jean-Karim Heriche]

The p-value and log2FC thresholds are entirely arbitrary. Just rank your genes in decreasing order of differential expression and take the top N, N being whatever is reasonable in your context. Sometimes, it may help to plot the values in decreasing order. You may notice an elbow which may suggest a convenient cut-off.

[1769mkc]

okay.I will try it, I have another query would you suggest to use Z score normalisation of the rna-seq data which are already normalised to plot heatmap ?

[Jean-Karim Heriche]

It depends on what your RNA seq data is. Search for related posts here on Biostars.