Differentially Expressed miRNA analysis : DEseq VS. Cuffdiff

Question

[illinois.ks]

Hello,

I am trying to identify differentially expressed microRNAs from my NGS data.

I have two conditions (control vs drug) Each condition has three replicates.

I have done the mapping procedure successfully and have done to identify differentially expressed miRNAs.

I have used three protocols.

1. DeSeq
2. edgeR
3. Cuffdiff

Interestingly, my result report different results especially for cuffdiff vs edgeR/DeSeq.

The result from edgeR/DeSeq looks very similar.

Under the qvalue < 0.05 threshold,

I have only 2 miRNAs by DESeq and 5 miRNAs by edgeR (luckily, those two miRNAs are overlapped.) Also, when I relax the q value threshold 0.1 , those 5 miRNAs are overlapped in both DESeq and edgeR

However, for Cuffdiff result, I have 27 miRNAs (including Mirlet7d, Mirlet7a-1 etc.. which were not detected by DESeq or edgeR). And among above 5 identified DE microRNA (by edgeR/DESeq), only 4 of them are overlapped with 27 miRNAs by Cuffdiff.. Apparently, one of them are not.

So, I am not sure which tool I have to trust. I know different method has different strategy to detect DEG. however, I am not sure for my study, which one I have to trust.

Could you please give some comments for this?

Answer 1

[Devon Ryan]

You need to validate the DE miRNAs in additional samples, likely with qPCR. Then you'll know which result set is more reliable.

[illinois.ks]

Could you please read the following posting?

Incosistent fold change signal : Differentially Expressed miRNA analysis : DEseq VS. Cuffdiff