Use of Irreproducible discovery rate (IDR) for RNA-Seq data

Question

[Sumit Paliwal]

I have done RNA-Seq for differential expression analysis. However, in certain cases, percentage of reads aligned varies significantly between the biological replicates. This is likely to influence the downstream analysis. Someone told me that I can use Irreproducible discovery rate (IDR) to see correlation between the replicates. If anybody has experience of using IDR in the context of RNA-seq please help me.

[Opt]

I've never seen it applied to that. Couple of issues I could see -- (1) IDR empirically estimates the CDF of the signal score for each replicate (signal score can be fold change, p-value, etc.). Chip-Seq has a much larger # of observations compared to gene expression tho so empirical estimate of CDF should be a lot better for Chip-Seq. (2) If you have 2 replicates and 2 controls -- are you going to generate 4 (2x2) lists of DE genes with p-values/fold change? Tho I guess you could combine controls and then test replicates against controls for DE.