Reporting all mapping combinations in SAM/BAM

[quinlana]

Hi all,
I am wondering if anyone is aware of an aligner that will report multiple mapping combinations for paired-end reads in SAM format. There are two cases that I am interested in:

1) when a pair maps concordantly to multiple loci, i'd like to report each such locus.
2) when a pair does not map concordantly to any locus, i'd like to report all mapping combinations between the two ends of the pair. that is, if end1 maps to one locus and the end2 maps to 10, i'd like to report all 10 combinations. if end1 maps 10 10 and end2 maps to 10, i'd like to report all 100.

now for case 2, i realize that the number of combinations can grow to huge numbers. therefore, it would be nice to choose up to N randomly-selected locations if the total number of possibilities exceeds N.

BWA chooses a combination randomly for cases 1 and 2. Also it seems that Novoalign merely reports the mappings for end1 and end2 separately without reporting the combinations. Mosaik may do this, but I am unsure.

Has anyone faced this issue and found an aligner that handles this? I have a program that generates all possible mapping combinations, but I'd prefer to have the aligner do this for me in SAM/BAM.

Best,
Aaron

[drio]

Try bfast. When postprocessing tell the software to report all the alignments. Once you have the sam (convert to bam) parse it and extract the information that you need (That's looking for those particular alignments you are interested on).
With this kind of postprocessing the sam output can grow a lot, but, if you split your input reads in chunks you should be able to handle it in parallel without problems.