DESeq2 for DE analysis of proteins

Question

[4732b159]

Hello, Does anyone use DESeq2 for differential expression of "proteins" (data is an expression count matrix of proteins), and have suggestions on whether suitability/limitations of the package etc.? Thanks!

Answer 1

[dsull]

What do your protein values look like? If they are intensity values rather than whole number values, why not use limma as many people do?

DESeq2 depends on whole number counts being negative binomially distributed.

[4732b159]

Thanks for your response, could you please elaborate what you mean?

[dsull]

If your protein "counts" look like:

Protein A: 0.4 Protein B: 4.583 Protein C: 3.582 Protein D: 2.492

Then you CAN'T use DESeq2 (see the decimals; they aren't whole numbers, they are CONTINUOUS numbers).

Just use the limma package for your DE analysis.

[4732b159]

Oh I see what you mean now, yes my data consists of whole numbers as expression counts.

[Antonio R. Franco]

DESeq2 requires the data being expressed according to a binomial negative.

[dsull]

Correct, and if it's whole numbers, check if it obeys that distribution; e.g. make a mean vs. variance plot.