Why Cuffdiff shows "no significant" expression difference in one of sample is 0 and another with a very high value?

Question

[Jingyue]

So here is the result from cuffdiff, I only printed out certain columns for read:

test_id    gene    value_1    value_2    significant
XLOC_226942     -       3.73672 0       yes
XLOC_227336     -       6.80194 0       yes
XLOC_227426     -       0       111172  no
XLOC_227811     -       0       154.721 no
XLOC_228955     -       0       9.80122 yes
XLOC_229214     -       6.61332 0       yes
XLOC_229754     -       8000.99 0       no
XLOC_231416     -       0       714.61  no
XLOC_231570     -       11.8838 0       yes
XLOC_231696     -       109.917 0       no
XLOC_231746     -       0       946.635 no
XLOC_231910     -       2.8676  0       yes
XLOC_231945     -       7.25977 0       yes
XLOC_232519     -       2.42438 0       yes
XLOC_232650     -       126.899 0       no
XLOC_232742     -       25.7937 0       yes
XLOC_233457     -       163.295 0       no

So why it shows no significant difference if one of value equal to 0 but another is over 100? Isn't that a huge different? And if one is 0 and another is less than 100, it considered significant. Do you guys have the explanation for this? Do I also need to consider those as significant expressed genes?

Thanks so much!

Answer 1

[Brian Bushnell]

This is a known issue with Cuffdiff, and one reason people generally prefer other programs (DESeq and edgeR) instead.