skewed cuffdiff result

Question

[kint1007]

Hi,

I have a question about using Cuffdiff with mouse single-end reads.

I get result (gene_exp.diff) skewed differential expressed genes (up: 113, down: 579).

That is weird.

So, I count the gene to check the whole expression is skewed.

      log2FC         FPKM == 0                        NOTEST  LOWDATA          |FC| < 1.2 (mid)                  yes
DOWN  < 0     9411   419 (Stress FPKM == 0 )  = 8992  2139    14       = 6839  1495              = 5344     ...  579
      == 0    4502   both 0
UP    > 0     10084  710 Inf(control FPKM=0)  = 9374  1523    26       = 7825  892               = 6933     ...  113

remove FPKM = 0, NOTEST, LOWDATA,

The number is similar. (6839, 7825)

Even after remove |FC| < 1.2, DOWN expression is 5344 and UP expression is 6933.

I wonder what reasons?

When cuffdiff calculate pvalue and qvalue, is there some filter option or something?

Thank you

Answer 1

[Ido Tamir]

You could try some diagnostic plots like MA or FPKM/FPKM or FPKM density across replicates etc, if you see a bias (with or without normalisation). Were the samples prepared in the same batch? If you don't see a bias and the samples were prepared in the same batch, and you see deregulation in mostly one direction - why do you think its biologically wrong?