FPKM values for DE analysis - Why?

Question

[Thalita]

Hello everyone!

I believe my questions are quite naive, but I am super new to RNA-seq data analysis, so forgive me! I already saw some questions regarding this subject, but I still do not understand it well. I am currently performing (trying to, actually) an analysis to identify differently expressed genes between two groups (with cancer, no normal samples in any group) from TCGA dataset. I've downloaded TCGA transcriptome data with FPKM-unstranded values, as the data is already normalized (is it right?!). However, when looking for a pipeline to perform this analysis, I noticed that DESEq2 or edgeR/limma analysis only uses raw counts as input to identify these genes. Why I should not be performing DE with FPKM values? I mean, which kind of analysis are FPKM values okay to perform with? Many many thanks in advance.

[Ming Tommy Tang]

if you want raw counts for DESeq2, follow this https://divingintogeneticsandgenomics.com/post/how-to-convert-raw-counts-to-tpm-for-tcga-data-and-make-a-heatmap-across-cancer-types/

Answer 1

[bk11]

I've downloaded TCGA transcriptome data with FPKM-unstranded values, as the data is already normalized (is it right?!).

Yes, these data are normalized.

Why I should not be performing DE with FPKM values?

While FPKM normalization method accounts for sequencing depth and gene length, FPKM data are not recommended for DE-analysis. The reason is that the normalized count values output by the FPKM method are not comparable between samples.

For your reference, I would like you to check this link.

[Thalita]

Guess I'm gonna adventure myself with the raw counts then. The link was super helpful btw, thank you!