Wierd issue with Cufflinks 2.2.0

[yueluo]

Hi all,
I'm having issues with results from cuffnorm.

I had 4samples(R1-R4) with 3replicates each, so that's 12 PE fastqs
R1= A1 A2 A3;
R2= B1 B2 B3;
R3= C1 C2 C3;
R4= D1 D2 D3

These were the steps I ran:
1. Tophat / Cufflinks for each fq.(get 12 BAMS + 12 transcript.gtf)
2. Cuffmerge to create merged.gtf.
3. Cuffquant to create cxb.(each BAM vs merged.gtf, get 12 abundance.cxb)
4. Cuffnorm with cxb and sample_sheet.
And got (TableA1)
I repeated step4, this time I treated replicates as individual samples.
And got (TableA2)

So now it seems cuffnorm re-arranged some of the columns, but FPKM values are the same.
Since this run was performed on a cluster where I had no root-rights, I used a pre-compiled version of cufflinks-2.2.0.

I then moved to our local server(with root-rights) and built cufflinks-2.2.0 from source.
I used the same reference/cxb/merged.gtf files to rerun cuffnorm with the same parameters above. And got (TableB1) and (TableB2)

To simplify things, all the tables where produced using '$> head genes.fpkm_table'. I also uploaded the "samples.table" of each run.
Seems to me that Tables A2 and B2 agree with each other, but obviously something's wrong with Tables A1 and B2.

Another thing I noticed : The pre-compiled version goes with SVN revision 4223, but the one I built from source is SVN revision 4222.

I wonder if anyone else has run into this problem?

[yueluo]

I've been having problems browsing Cufflinks' Google Group. But my friend helped post it:

https://groups.google.com/forum/#!topic/tuxedo-tools-users/eOsiM3rbrYE

Hope this doesn't cause any convenience.