bwa mem on a bam file

Question

[vimartin]

Hi,

I'm trying to use bwa mem for the first time and I wonder, is it possible to run it on a bam file instead of a fastq file? I noticed that bwa aln can be runned on bam file using '-b' but this option is invalid for bwa mem. The manual does not mention whether it is possible or not. I tried to simply run bwa mem on bam file and the output seems to be a header only sam file.

Does anyone has experience on the question?

Thanks in advance.

Answer 1

[vimartin]

Ok, I should have wait I guess. Finally I found the answer which is : no. This was detailed in the mail archive announcing the release of bwa mem. I put it as an answer in case someone else ask himself the same question.

http://sourceforge.net/p/bio-bwa/mailman/message/30527151/

[Jorge Amigo]

you could try piping samtools fastq (former samtools bam2fq) to bwa and save both disk and time:

samtools fastq reads.bam | bwa mem ref.fa -

note: this works only for non-paired-end reads.

[rgiannico]

Nice solution, but are you sure the -p parameter works well? Does bwa correctly pairs the paired end if they come from the stdin ? Have you tested it?

[Jorge Amigo]

from bwa manual page:

If -p is used, the command assumes the 2i-th and the (2i+1)-th read in reads.fq constitute a read pair (such input file is said to be interleaved)

but unfortunately samtools fastq does not output interleaved reads.

you can either try sorting reads by name BEFORE fastq generation:

samtools sort -n reads.bam | samtools fastq - | bwa mem -p ref.fa -

or you can either try sorting reads by name AFTER fastq generation (which is slightly faster):

samtools fastq reads.bam | paste - - - - | sort -k1,1 -S 3G | tr '\t' '\n' | bwa mem -p ref.fa -

[abcoxyzide]

I found this suggested solution from samtools' documentation and modified it according to our context:

samtools collate -u -O reads.bam | samtools fastq -0 /dev/null -s /dev/null - | bwa mem -p ref.fa -

samtools collate is probably much faster than sorting by name. I haven't tested it myself though, so future comers should use it with caution. Also, this code discards the singleton / supplementary / secondary reads, which may or may not be desired.

[Nicolas Rosewick]

you could use samtofastq from Picard to convert your bam to fastq prior BWA mem alignment.