what is the best p value cuttoff to select differentially expressed genes ?

Question

[jack]

Hi,

I have don gene differential expression analysis of RNA-seq data using Deseq2. it's report the list of the differentially expressed genes with Fold change, pvalue and adjasuted P-value.

I know that, I should stick with adjusted P-value. but:

1. What is the good and trust able threshold for P-value?
2. Why just looking at fold change can be misleading?

Answer 1

[Devon Ryan]

1. 0.1 is a common threshold for adjusted p-values
2. suppose you have an outlier or a lowly expressed gene such that the observed fold-change is questionable...

For 2, the reality is that DEseq2 moderates the fold-change, so this should then be a bit more useful. still, at the end of the day you're not just interested in the magnitude of change between group but also how reliable that is.

[jack]

thanks. but you don't think 0.1 is not very low threshould ? you mean 0.01 ?

[Devon Ryan]

I meant 0.1, 0.01 would be overly restrictive for an adjusted p-value. Of course, the threshold you'd actually use is dependent on what you're doing downstream. If you're going to base an expensive clinical trial on the results, then you'd use a more restrictive threshold.