Tophat/Cufflinks treating paired end reads as single

[Helical]

I am using paired end reads with Tophat/Cufflinks and although I get great mapping with Tophat and a decent number of novel transcripts with cufflinks I think the software is not treating my reads as paired end. I suspect I have some minor formatting problem with my tophat command but maybe it’s more complicated. My commands:

tophat -i 20 -I 100000 -m 1 -g 1 –tophatrun_1 -p 2 -r 180 --transcriptome-index=known/known_transcripts Gen/Genome A01_1.fastq A01_2.fastq

I get >90% mapping of the reads here, but when I go to use cufflinks with this command:

cufflinks -I 100000 --min-intron-length 20 -g transcripts.gff -p 2 -o cufflinksrun_1 tophatrun_1/accepted_hits.bam

I get the following error:

“Warning: Using default Gaussian distribution due to insufficient paired-end reads in open ranges. It is recommended that correct parameters (--frag-len-mean and --frag-len-std-dev) be provided”

The run completes, but this implies to me that although Tophat is mapping my reads fine, it is treating them as 2 pools of single end reads and not as paired end. Otherwise why would cufflinks be unable to determine the fragment length? Is my formatting wrong somewhere? Thanks for any help you can provide!


<EDIT> I've inspected the accepted_hits.bam file using SAMTOOLS and it looks like >90% of the reads ARE in fact being mapped as a pair. I am really at a loss as to why Cufflinks is giving an error about "insufficient paired end reads in open ranges".

[rflrob]

How many reads do you have? I'm not certain how many cufflinks wants before it will estimate them, but it's somewhere under a million.

[Helical]

Hi rflrob,

Thanks for replying. I have about 30 million reads per library so I don't think that is the problem.

It's strange that it is even trying to look for "open ranges" since the manual states it only does this without a reference transcriptome and I am supplying one.

[tir_al]

Dear Helical,

It seem I have bumped into the same problem that is troubling you - cufflinks recognizes paired reads as single end ones.
Did you manage to solve your problem?

My best guess is that either the read names are not properly formatted (e.g. _1/_2 or some kind of suffix is missing), or there is a missing tag that cufflinks expects to see.

Best!

[jp.]

Hi All
can someone tell me how to give commands for read length 101 x 2, insert size 80-380(main 150) Hiseq2000-PE ?

[dekontrol]

Same problem. Moreover, Raw Map Mass becomes very very little respect to Normalized Map Mass. Does anybody found a solution?

Thanks