In order to find out whether your gene upregulated, you can consider log2 fold change greater than 2 having both p-value and q-value < 0.05, however you can go upto fold change greater than 1. Further if you want to identify the Expression of particular upregulated genes, you can calculate their expression values (RPKM, FPKM or TPM) using R statistical package.

I think you should make a critical distinction between your inferences on the one hand, and the fold changes (taking into account confidence intervals of some sort) on the other hand.

If everything is done correctly, your q-values (or another measure of statistical significance like posterior probabilities and the likes) is the first thing to look at. You decide how "certain" you want to be about your conclusions. For example: if you plan to use these data for future research, it's unwise to bee too stringent, because you will lose interesting information, and your false positives are not a drama because they will surface in future experiments. On the other hand, if you want to design a bio-assay or a diagnostical test based directly on these data, you better be super certain...

After you have decided on your desired level of confidence, you can look at the fold changes. Here again, it totally depends on what you want to do with it. For mere research purposes, don't cut it off: any significant change may be interesting, and also bear in mind that a fold change in mRNA levels does not directly correspond to a similar fold change in translated protein, or whatever effect your siRNA has in the cell in your case. The more you cut-off, the less info you will retain for downstream analysis like pathways etc.

However, again, if you want to turn this into a diagnostic test, it will be a lot more straightforward to be looking at things that change tenfold, than to things that change by 20%.

Hope this helps

Best,

Rik

Thanks all for your comments. 

The nature of these databases is the following:

(a) 12,600 genes included, genes p