Convert SRA to FASTQ with fastq-dump but problem of read length

[MurielGB]

Hello,
I have Illumina paired end reads of length 76bp.
The problem is that when I use fastq-dump to obtain two files with paired reads separated, it splits the reads into 101bp and 51bp rather that 76+76...
I tried with the options --split-files, --split-spot, --split-3 and always have the same result.
I also tried different fastq-dump versions: 1 ; 2 ; 2.3.4 and 2.3.5.2.
Do you have an idea how I can do that ?
Thanks !

[GenoMax]

It is possible that the dataset you are looking at has asymmetric reads. Have you looked at the record in SRA to see if that is the case?

[MurielGB]

I don't know if this is possible.
When I convert SRA to FASTQ without any option, I obtain a fastq with 152bp reads.

Here is the page where I downloaded the sra file : http://www.ncbi.nlm.nih.gov/sra/?term=SRR1174239

[GenoMax]

Based on the record it looks like a standard 2 x 101 bp PE dataset.

Update: The information in SRA appears to be incorrect since the dataset is dumping with 152 bp length (so would be 2 x 76).

[MurielGB]

Yeah I agree but then why do I obtain 152bp reads when using fastq-dump ?!!

[GenoMax]

This appears to be an asymetric dataset (101 x 51) as originally suspected. See attached screencap.

[MurielGB]

OK but when I do fastqc on the fastq file with 101bp reads, I obtain the attached graph that is, to me, typical of problems of read splitting with bad length.

[GenoMax]

It does appear that something is fishy.

You may want to email SRA support and ask them to look into this data set. You could also alert the submitter independently.

[MurielGB]

OK, thanks a lot !

[aaronh]

I ran into this issue with another data set. The problem was SRA miss-parsed the fastq files. A few emails between the help desk and the original depositor resulted in SRA reformatting the files.