Read Counts Difference From Different Tools

Question

[camelbbs]

Hi, I got reads counts from rnaseq from different tools, I am not sure why the cufflinks gives so different counts result. Anyone have experience on that?

    ACTB    40264.3    38071.2    26648.7    41528.2    39006.5    40361.8    47723.2    25179.8   cufflinks
    ACTB    2889    1102    1793    2264    2747    1836    624    2316   htseq
    ACTB    2084    875    1362    1719    1919    1345    475    1722    partek

Thanks, Cam

[Sukhi Singh]

Check this post Deseq, Edger And Cuffdiff - Different Result, might give some hints :)

[Neilfws]

Is that the correct link? I do not see that phrase, nor anything which answers this question.

[Sukhi Singh]

now corrected !!

[Hayssam]

I think it was about differential expression, and that the problems were eventually with cuffdiff, not cufflinks. I think Cam's problem is at the read count estimation stage.

[Hayssam]

You can check these links:

• Inconsistency Between Number Of Genes In Cuffdiff And Htseq
• http://seqanswers.com/forums/showthread.php?t=9129

How did you run cufflinks? Its estimates are orders of magnitudes higher than the other tools.

[camelbbs]

Thanks, I run cuffdiff like this: "cuffdiff -p 8 --uper-quartile-norm -u -b hg19.fa -o output merged.gtf s1.bam s2.bam s3.bam ..."

[Hayssam]

Cam, could you try to run cuffdiff without the quartile normalizatioin?

[camelbbs]

Hi massyah, I tried with no --uper-quartile-norm option and got the same result... So I think the problem might not be here..

[Fabio Marroni]

Could it be that cufflinks returns FPKM and htseq and partek return the other two read counts? I just read the htseq man page here http://www-huber.embl.de/users/anders/HTSeq/doc/count.html#count (I never used the software so ... beware!) and I think I understood that the software actually count the reads mapping, and not FPKM. I don't know what partek does. You can track read counts in cuffdiff output files such as "genes.count_tracking", but I never didi it.