Cuffdiff Question with FPKM of one sample being 0

[Zapages]

Hi Everyone,

I have question in regards to cuffdiff. I have the following data in tissue 1 vs tissue 2.

Sample A: 0 Sample B: 20.292 Log2 Change(fold change): 1.79769e+308

My understanding for the fold change to be 1.79769e+308 is due to the cufflinks/cuffdiff normalization method goes to close to infinite due to sample A having 0 FPKM value

But what other explanation could there be for this?

Many thanks in advance.

[kmcarr]

This is just an artifact of the numeric type which cufflinks uses to store the Log2FC value. Apparently it is storing this value as a double because the largest possible value for that type is 1.79769e+308.

[Zapages]

That is interesting, but should we take this FPKM value for this number as something to be highly expressed or should we take a look at isoforms and see the confidence intervals to consider its true expression.

I just found two interesting discussions about it, where they state the FPKM should not be taken for the face value and the confidence intervals seriously consider for gene to be expressed.

https://www.biostars.org/p/72486/ , https://biostar.usegalaxy.org/p/5022/ , and https://biostar.usegalaxy.org/p/4940/


I was wondering what is your opinion about this.

Thank you again.