How to obtain RPKM/FPKM from HTSeq-Count data?

Question

[alcs417]

Hi there,

I am currently working on a project related to pan-cancer analysis. I have done differential expression analysis for miRNAs and genes with edgeR. I know that edgeR only takes raw counts as input, so I downloaded the HTSeq-Count data from GDC data portal. After the differential expression analysis, I'd like to obtain the normalized expression values of miRNA/genes (RPKM/FPKM) for the downstream analysis, such as using pearson correlation between miRNAs and mRNAs to construct a miRNA-mRNA regulatory networks and so on. However, I got stuck for days on how to get the normalized expression values of both miRNAs and genes. Here are my questions:

1. There is a function called "cpm"(counts per million) in edgeR, but it says it doesn't take the gene length into accout; edgeR also provides another version of normalized counts "pseodu.counts", however, someone says this is quite difficult to interpret. So I am wondering if I could use "logCPM" as the normalized expression values for the downstream analysis?

2. If not, I realized that there is also a function called "rpkm" in edgeR which could calculate the normalized expression values. However, it needs the gene/microRNA length information to make it work. I do not know where to find the length information for genes and microRNAs, since there is no such information contained in the HTSeq-Count file. Could any one please tell me how to do it? Should I download the gene information from ENSEMBLE and the miRNA information from mirbase? And calculate the length information by myself? Is there any R package that could do the work instead?

3. Could I just download the RPKM files of miRNAs and genes from GDC data portal to construct the miRNA-mRNA regulatory network? Is that right? It seems to be the easiest way for me to do though....

Any help would be really appreciated.

Answer 1

[GouthamAtla]

There is no much use of calculating RPKM/FPKM for miRNAs.

To answer your question, you can use featureCounts to quantify your genes using a GTF file. This outputs a matrix of counts, which also includes a column of gene lengths. This column can be given to edgeR rpkm() function.

Option 3 would make sense. If you already have normalised data available, you can use to calculate the correlations.

[alcs417]

Could you please tell me where to download the GTF file? It seems that TCGA does not have this file for genes? Also, how can I obtain normalized miRNA expression data from HTSeq-Count data then? Thanks.

[GouthamAtla]

I assumed that you already have the RPKM from this "Could I just download the RPKM files of miRNAs and genes from GDC data portal"

I would suggest to speak to someone in your workplace who does bioinformatics. If you are starting with RNA-Seq analysis and wants to work with TCGA, it needs lot of work and guidance.

[alcs417]

I eventually found the GTF file from the ensembl website. Anyway, really thanks for your help. I am working on it.