Can anova analysis be done on one vs many microRNA count data

Question

[Saad Khan]

Hi,

I have one cancer samples and 4 normal samples for microRNA in humans. I was wondering if anybody has done some sort of differential expression analysis using similar techniques like ANOVA e.t.c for such type of experimental design.

Answer 1

[Devon Ryan]

Well, you can use an ANOVA-like procedure, but given that ANOVAs are meant for continuous measurements and you have discrete counts an ANOVA itself isn't the best option. DESeq2, edgeR, etc. can all handle data like this.

[Saad Khan]

Deseq2 does not talk about such design in its manual/vignette. Can you point me to specifc place where such an experimental design is handled. Where one sample has no replicate while other has many.

[Devon Ryan]

It's not explicitly mentioned that I recall, but a design like this will work with any GLM-based package. The only caveat is that the fold-change may be a bit underestimated, since the estimate from the single observation is going to be very heavily influenced by the prior.

BTW, at least the original DESeq mentioned experiments like this. They're not as common anymore, which is why fewer people mention them.