Determine Top 100 Differentially Expressed Genes

[wilson90]

Dear all,

I am wondering if we decide the top 100 differentially expressed genes based on:
A) Log Fold Change or
B) P-value
C) Or a combination them? (In the case, order matters? (logFC first, p-value?)

Thank you.

Best Regards,
Wilson

[sisch]

I'm not sure that it makes sense to look only at the top hundred of differentially expressed genes from a statistical point of view. Either they are significantly different or not.
However, if you want to sort your data a bit, I would suggest the following:

1. Determine differential expression by statistics (whatever test you see fit)
2. From the set of significantly differing genes, sort by fold-change (or log-fold-change)

This way you get a list that has on top the most differing (in terms of expression) genes that are confirmed to be differential by statistics.

Still, looking only at a subset of results that came out of a statistical test as significant feels wrong.