Consistent FPKM counts

Question

[pezhmansafdari]

Hi,

I have RNASeq data from 8 different cryosections. I need FPKM counts to do gene clustering. When i use cufflinks to get FPKM counts it does not cover the whole 31000 genes and even different number of genes are covered for different cryosections. Is there a way to force cufflinks to cover all the genes, even if the count is zero, or is there any other software for doing this?

Thanks a lot,

Pezhman

Answer 1

[Asaf]

You can use raw counts for clustering so htseq-count should work for you.

[Devon Ryan]

It should be noted that you actually use the variance stabilized or robust log transformed normalized counts, rather than the raw counts, for clustering.

Answer 2

[GouthamAtla]

As far as I know, you will zero FPKMs as well. But if your goal is to cluster the data, you could use htseq-count as suggested by Asaf, and then you will have raw counts. If you want normalised counts, use DESeq2 or edgeR for normalisation and then do clustering. But it is always good to use normalised counts for heatmaps/clustering.