converting paired-end (PE) bam file to single-end (SE) fastq

[adrian]

Hi:
while working with COAD TCGA BAM files, I find the very annoying to find PE reads. These files are mashed up and not consistent.
for example:
1. read lengths are not consistent. Some are 34 some 76 reads.
2. Many reads miss mate or pair.

I want to identify novel splicing differences however TCGA BAM files are mapped to known transcripts (known exon pairing from known isoforms gtf) thus limiting the discovery of novel isoforms.

I decided convert BAM to fastq and realign to full genome.

While doing this, because of loss of many pair and mates in bam, I converted them to single end fastq.

Any ideas if converting a paired-end bam to single end fastq pose any problem in philosophical ways.

thanks

[dpryan]

Yes, you'll be expected to decrease your mapping efficiency a bit, since one mate can act as an anchor to rescue the other. Further, it's much easier to use paired-end reads to find isoforms, since you're then not relying solely on alignments over a splice junction.

[adrian]

Yes thats a disadvantage I agree.

Unfortunately, the bam file does not have enough PE reads.

When I used bamtofastq for PE fastq files, interestingly I obtained 0 fastq reads.

[Brian Bushnell]

Adrian,

You can try running repair.sh to split the file into paired and unpaired reads, and then map twice, once for the paired and once for the unpaired, and then merge the bam files. That will allow maximal use of the available information.