Cuffdiff results puzzling me. fpkm 350 vs 5.6 IS sig while 6 v 350 ISN'T ??

[beeman]

Hi all,

New to RNAseq analysis and trying to find DEGs with some PE seq data I've aligned using tophat. I just can't understand why cuffdiff is calling a gene diferentially expressed in one comparison and not the other.

I ran cuffdiff against 3 samples and in the gene_exp.diff file, the FPKM values in comparison 1 are: 350.999 & 5.63659 which are found to be significant (duh!).. But in the comparison between samples 3 and 1 with FPKM values of 6.01634 & 350.999 the gene is not considered to be significantly different.

I assume this is due to the low FPKM of this gene in sample 3, but the status call is OK in this analysis... Can someone please explain why this is happening? I really can't understand it.

Data in question:
Comaprison 1. ( sig )
ACYPI004698 ACYPI004698 acyp2eg0000333 chr1:4083377-4092062 alatae aptarae OK 350.999 5.63659 -5.9605 -6.2077 5e-05 0.0233079 yes
Comparison 2. ( NS )
ACYPI004698 ACYPI004698 acyp2eg0000333 chr1:4083377-4092062 oviparae alatae OK 6.01634 350.999 5.86644 6.34932 0.0005 0.0919123 no

Thanks !!

[dpryan]

It'll depend on the variance in each case. It looks like that of the second comparison is higher (judging from the p-value). Also and adjusted p-value of 0.09 is normally considered significant (the normal threshold is 0.1, which I know is different from the standard 0.05 for unadjusted values), particularly since the more recent versions of cuffdiff seem to be quite conservative.