edgeR analysis outputs FDR equals 1

Question

[silas008]

Hello, people!

I have found different answers for similar questions but none of them fits with my case. I am using edgeR to do the statistics of my RNAseq data and, as shown bellow, I have 3 replicates for each of two groups:

> x <- read.delim("youngxsenes_comparison_table.csv",row.names="miRNAs")
> group <- factor(c(1,1,1,2,2,2))
> y <- DGEList(counts=x,group=group)
> y <- calcNormFactors(y)
> design <- model.matrix(~group)
> y <- estimateDisp(y,design)
> et <- exactTest(y)
> topTags(et)


                    logFC     logCPM      PValue FDR
hsa-mir-3613    0.9061035  8.5155819 0.001912410   1
hsa-mir-6825    2.3963572 -1.6642635 0.001941057   1
hsa-mir-133a-1 -3.3808103  1.9433720 0.002665153   1
hsa-mir-133a-2 -3.0362991  3.3413262 0.003145797   1
hsa-mir-5008    1.9878428 -0.7387409 0.003295906   1
hsa-mir-214    -0.6672389 11.3911591 0.005099100   1
hsa-mir-124-1   1.3643040  1.5929070 0.005458540   1
hsa-mir-124-3   1.2313130  2.6963757 0.005794421   1
hsa-mir-203a   -2.1373892  2.5981025 0.005921462   1
hsa-mir-522    -2.7536418 -1.9466949 0.006955708   1

Even though my data have 3 replicates, FDR remains equal 1. I read this is common for data with no replicates. Does anynone know why this also happens in my case?

Thank you

Answer 1

[swbarnes2]

The link below might explain it...see how a p-value of 0.05 isn't good enough when you have 20 tests? If you have a 500 genes, a p-value of 0.002 isn't good enough. You might really have no real changes, or those changes might be too subtle to show with only three replicates.

https://xkcd.com/882/