Question: Rna Raw Count and Gene expression correlation

Question

[salvatore.digiorgio]

Hi.

Starting From TCGA dataset HTSeq - Counts I used TCGAanalyze Normalization (TCGAbiolinks package) based on EdgeR to obtain normalized count matrix. On these data can i perform correlation test or Is better start from HTSeq - FPKM? Thank you.

[Jake Warner]

Could you provide a little more info on your design? What are you hoping to correlate? If you wanted to find the most highly expressed genes for example then using the normalized count matrix will suffice. If you are looking to compare expression profiles between genes across a series of samples then an additional standardization or transformation would help.

I also agree with Wouter that your edgeR normalization is better than FPKM.

[salvatore.digiorgio]

I'm trying to find correlation of expression levels of a small set of genes belonging to one type of cancer with different sample set, but not across different types Cancers sample set.I hope I made myself clear, What be the right course?

Answer 1

[lessismore]

It's much better to start from normalized data. FPKM which take in consideration the library size and the gene length.

[WouterDeCoster]

But OP generated normalized counts using edgeR, which is a more sophisticated method that FPKM. Wouldn't edgeR normalization be superior to a simple FPKM transformation?