Can FPKM data sets be of any use or are they trash?

Question

[schulpen_91]

Hello all. I have acquired FPKM-normalized data sets (excel files) which are (supposedly) to be used for differential expression analysis.

As is well known, FPKM normalization is not the best strategy in the current day and age. It can't actually be used for DE, and it can't be properly imported into popular tools such as DEseq2, EdgeR, and Limma, because the results would become rubbish. The original count files are not available as far as I know so that's problematic.

Is there any way to make these data sets useful and perform a proper DE analysis or am I absolutely doomed?

[Istvan Albert]

As per the link cited by jared.andrews07 below, you can still use limma

If FPKM is really all you have, then convert the values to a log2 scale (y = log2(FPKM+0.1) say) and do an ordinary limma analysis

Answer 1

[i.sudbery]

I'd note that there are perfectly valid uses for data on an FPKM scale (although often TPM is better). However DE analysis is not one of them.

Length normalised data is useful when you want to compare the expression of different genes within a sample, not compare the expression of a gene between samples.

Answer 2

[jared.andrews07]

Pretty doomed. See: How to do FPKM differential analysis? https://support.bioconductor.org/p/56275/#56299

Your time is certainly better spent elsewhere.