Hi,

We have been applying lfcThreshold test in DESeq2 with altHypotesis=lessAbs to find what microRNAare significantly stably expressed between matched fresh frozen and formalin-fixated paraffin-embedded tissue. In the official manual, they use lfcThreshold=0.5. When we use the same threshold, we don't get any significantly stable miRNAs using the BH-adjusted p-value (and just a few for non-adjusted p-value). We have also tried with lfcThreshold=0.75 and lfcThreshold=1 with obviously better results.

My question is: What can be considered a good lfcThreshold-value for finding stably expressed microRNAs?

As far as I can see the significance is calculated from baseMean and it's standard error.

Best Regards,

Maria

The value set in the manual is just an arbitrary value. You should pick a fold change which is meaningful for your biological system and then take the log2 of that fold change and supply that to lfcThreshold. E.g., if you care about more than doubling, use log2(2) = 1. You could also use the MA plot to inform your threshold choice.