Minimum number of transcripts in cufflinks GTF

[lfx]

Hi all,

Would like to know if there is a minimum number of transcripts required in a GTF file in order for cufflinks to run properly?

I've also tried this with cuffdiff:-
1) full genome GTF
2) take 10 transcripts out of the full GTF

the FPKMs seem to pile on in the results from the smaller GTF file.

Am I doing something wrong? The reason I'm doing this is for screening some potentially novel transcripts / locations in the genome, and would likely be working with a much smaller set of GTF entries.

Advise is much welcomed!!!

[adarob]

FPKM is Fragments Per Kilobase per Million reads sequenced. Currently, Cufflinks calculates this "million reads sequenced" as "million reads mapped to the annotation", which is something we are looking at switching in the next version. To put the FPKMs on the same scale for the different runs, simply multiply the FPKMs by the "Total Map Mass" that Cufflinks prints to the screen. You can then divide by the number of reads from the full data and everything will be on the same scale.

[lfx]

adarob, thank you for the clarification.

does this mean the cuffdiff results have already taken this into account when computing fold change between 2 samples?

because if each run has its own map mass and number of reads, then i would think that their fpkms will not be on the same scale.

[adarob]

Since they both are forced to use the same GTF in cuffdiff, they will be on the same scale.