no cuffcompare .stats with more than 6 samples

[sudders]

Hi All,

I'm having a weird issue with the output from cuffcompare, and I wondered if anyone had seen the same.

We are using cuffcompare to comapre and track transcripts build from an RNAseq project. We use the following commandline to run cuffcompare:
cuffcompare

  cuffcompare -o example.cuffcompare 
                    -s /ifs/mirror/genomes/bowtie/hg19.fa
                     -r <( gunzip < refcoding.gtf.gz)  
                   ./ctmpL7zs9K/example1.gtf.gz
                   ./ctmpL7zs9K/example2.gtf.gz &> example.cuffcompare.log

According to the cuffcompare manual, cuffcompare is supposed to output a <outprefix>.stats file. We never get this file, although usually we do get the stats in a file name just <outprefix> (e.g. in this case example.cuffcompare). However, this only happens when there are 6 samples or fewer. If we use 7 or more samples, not stats are output anywhere that we can find.

I'm running cuffcompare 2.0.2 using precompiled binaries on 64bit linux. But i've tried with 2.0.0, 1.4.0 and 1.3.0. Same result each time.

Anyone else seen this? We'd really quite like the stats as our pipeline uses them for quality control.

Cheers,

Ian
---

[westerman]

Are you talking about something that looks like the following output file? I deleted some of the following because the project is still on-going but you should be able to see that I have over 6 samples.

# Cuffcompare v2.0.2 | Command line was:
#cuffcompare -o proj123 -r westerm/(deleted)/References/genes.gtf ../001664/tophat_with_gtf/transcripts.gtf ../001665/tophat_with_gtf/transcripts.gtf ../001666/tophat_with_gtf/transcripts.gtf ../001667/tophat_with_gtf/transcripts.gtf ../001668/tophat_with_gtf/transcripts.gtf ../001669/tophat_with_gtf/transcripts.gtf ../001670/tophat_with_gtf/transcripts.gtf ../001671/tophat_with_gtf/transcripts.gtf ../001672/tophat_with_gtf/transcripts.gtf ../001673/tophat_with_gtf/transcripts.gtf 
#

 Total union super-loci across all input datasets: 109817 
  (12469 multi-transcript, ~1.2 transcripts per locus)

[sudders]

Yes, exactly.

One would normally expect to see a whole load of stats about the run,

something like:

# Cuffcompare v2.0.2 | Command line was:
#cuffcompare -o tax-Pre-agg.cuffcompare -s /ifs/mirror/genomes/bowtie/hg19.fa -r /dev/fd/63 ./ctmpL7zs9K/tax-Pre-R1.gtf.gz ./ctmpL7zs9K/tax-Pre-R2.gtf.gz ./ctmpL7zs9K/tax-Pre-R3.gtf.gz ./ctmpL7zs9K/tax-Pre-R4.gtf.gz ./ctmpL7zs9K/tax-Pre-R5.gtf.gz ./ctmpL7zs9K/tax-Pre-R6.gtf.gz
#

#= Summary for dataset: ./ctmpL7zs9K/tax-Pre-R1.gtf.gz :
#     Query mRNAs :  116810 in  110584 loci  (21299 multi-exon transcripts)
#            (5204 multi-transcript loci, ~1.1 transcripts per locus)
# Reference mRNAs :   76280 in   19285 loci  (74791 multi-exon)
# Corresponding super-loci:          13596
#--------------------|   Sn   |  Sp   |  fSn |  fSp    
        Base level: 	 43.3	 24.2	  - 	  -                    
        Exon level: 	 24.1	 36.5	 26.6	 40.3                 
      Intron level: 	 37.5	 95.3	 38.1	 96.7      
Intron chain level: 	  6.8	 23.8	 12.7	 44.6
  Transcript level: 	  0.0	  0.0	  0.1	  0.0
       Locus level: 	 23.8	  4.2	 29.7	  5.3

Matching intron chains:    5073
         Matching loci:    4584

          Missed exons:  152593/320767	( 47.6%)
           Novel exons:   90245/211449	( 42.7%)
        Missed introns:  122705/246111	( 49.9%)
         Novel introns:    2149/96936	(  2.2%)
           Missed loci:    5528/19285	( 28.7%)
            Novel loci:   60885/110584	( 55.1%)

[COLOR="Red"][B][I]############<SNIP> one table like this for each sample </SNIP>###########[/I][/B][/COLOR]

 Total union super-loci across all input datasets: 191274 
  (11845 multi-transcript, ~2.7 transcripts per locus)

[westerman]

I generally do not look at the stats file -- although you've convinced me that I should do so more often. Looking at my projects over the past several months reveals the following. Each project section has the project name, the cuffcompare version, the number of samples in the project and the number of 'Base level' lines in the stat file. If this latter number is the same as the number of samples then a full report was generated.

Project #1
v1.3.0
9
9
Project #2
v2.0.0
5
5
Project #3
v1.3.0
6
6
Project #4
v2.0.2
12
0
Project #5
v2.0.2
12
0
Project #6
v2.0.2
6
6
Project #7
v2.0.2
12
0
Project #8
v2.0.2
8
0
Project #9
v2.0.2
12
0
Project #10
v2.0.2
14
0

So, to me, it does look like a change in cuffcompare v.2.0.2. However I only have one pre-2.0.2 and greater than 6 sample project to back up this statement. Also you said you have tested this with other versions ... so ... I don't know what is happening. I do see the same effect as you do, at least for v.2.0.2.

[mondongho]

If we use 7 or more samples, not stats are output anywhere that we can find.

---

Did you ever find the error/solution?

[sudders]

No sorry, i've just given up using the provided stats on such large projects.