Manta and alignment name collision

Question

[German.M.Demidov]

Dear community members,

I received hundreds of CRAM files which I have to run through Manta SV calling and they fail due to "Unexpected alignment name collision" - this file contains tens (out of millions) of reads which were multi-mapped, so they have 2 entries for some reads with particular IDs.

I don't care about those tens of multi-mappers, I want to keep only one alignment out of two (the random one) - what's the cheapest way to do it? Previously for WES samples I was converting bam files to fastq, realigning and removing reads with same IDs, but now I have WGS and I certainly don't want to re-map them.

If anyone has a ready-to-use script - please, share, otherwise I'll have to do something in python/samtools view myself...

UPD: Temporary solution - I do samtools view, pipe into 5-liner in Python and then pipe back to samtools view and this is the weirdest routine I've ever done in bioinformatics

Answer 1

[Pierre Lindenbaum]

I wrote https://jvarkit.readthedocs.io/en/latest/SamRemoveDuplicatedNames/ to remove this kind of reads, but I'm not sure it's what you're looking for.

another idea: change manta to skip those reads. for example, replace https://github.com/Illumina/manta/blob/75b5c38d4fcd2f6961197b28a41eb61856f2d976/src/c%2B%2B/lib/manta/SVCandidateSetData.cpp#L125 with 'return;` instead of throwing an exception (not tested !!)

[German.M.Demidov]

Thanks a lot! Will try both!