bam to paired AND unpaired fastq reads

[Kennels]

Hi,

Some of the posts I found here are a bit old, so in case there new tools...

Does anyone know of a tool that extracts from a .bam file into both the paired reads fastq AND unpaired (single) reads fastq?
Picard SamtoFastq does not do this, and a tool from bamUtils does it but it is quite slow.

I don't care about mapping information, I just want the paired and orphaned reads.

Thank you

[dpryan]

If you're familiar with programming, the samtools C API is pretty convenient and could be used to do this quickly. There's also pysam in python, which I assume is a bit more approachable than using C (I've never used it and don't know what your background is).

[kmcarr]

Try piping samtools and picard together. Using the filtering flags for samtools view you can output only single or paired reads and strem those to picard SamToFastq.

Paired end

samtools view -uf 1 input.bam | java -jar SamToFastq INPUT=/dev/stdin [rest of options]

For single end change the 'f' to 'F' in the samtools command.

*This solution is untested

[sBeier]

If you filter your bamfile to get a) only the mappings where both reads map and b) the others you could use bedtools bamtofastq. It's not fast, but you can get both paired fastq for the first file and single ones for the second.

[Kennels]

thanks everyone.
I tried kmcarr's solution, but ran into errors due to some mates not being present in the .bam.

I found this solution: http://seqanswers.com/forums/showthread.php?t=16395

and eventually was able to compile the code, and it works very nicely and quickly (under 5 mins for a 2.5Gb bam). It outputs both the paired reads and orphan'ed reads, but found that it will also duplicate a read if there are multiple alignments for it.

Since my goal was to just get the reads that could simply align, I regenerated the .bam (accepted_hits.bam from tophat2 output) to report only one alignment per read.