Delly shouln't use INFO/END for the BND / TRANSLOC

[lindenb]

Hi,
unless I'm wrong Delly2 uses the INFO/END attribute to specifiy the second position of the breakpoint. This breaks the integrity of the tabix/bcftools indexes

See

samtools/hts-specs#266 (comment)

and

macarthur-lab/gnomad_browser#140

Thanks

[tobiasrausch]

Hi,

Indeed, since 8 years Delly uses CHROM:POS with INFO/CIPOS for the first breakpoint and INFO/CHR2:INFO/END with INFO/CIEND for the second breakpoint. From an SV point of view that's very consistent because every paired-end/split-read cluster gives rise to 2 breakpoints and INFO:CT clarifies how the two ends are joined and that works for both, intra-chromosomal SVs and inter-chromosomal SVs. I guess, INFO/END and INFO/CIEND don't create issues for intra-chromosomal SVs for tabix/bcftools? I am happy to change the notation for inter-chromosomal SVs but then the specification should clarify how the second chromosome is reported (INFO/CHR2 ?), the second breakpoint position (INFO/POS2 ?) and the confidence interval (INFO/CIPOS2 ?). As previously discussed in this issue I am very reluctant to break-up inter-chromosomal translocations into multiple records because this notion of having one VCF record for each paired-end/split-read cluster is easy to work with and prevents duplicating genotypes across hundreds of samples.

Best, Tobias

[tobiasrausch]

Hi,

Delly uses now INFO/CHR2 and INFO/POS2 for inter-chromosomal translocations to be VCF-compliant. For all intra-chromosomal SVs POS and INFO/END is used.

Thanks, Tobias

[fgvieira]

Hi Tobias,

are these fixed supposed to be already on version 0.8.2? I used it on a sample and got this SV:

2       33141291        BND00002717     T       [1:19212036[T   .       LowQual IMPRECISE;SVTYPE=BND;SVMETHOD=EMBL.DELLYv0.8.2;CHR2=1;END=19212036;PE=2;MAPQ=33;CT=5to5;CIPOS=-460,460;CIEND=-460,460

but tabix complained that the coordinates were wrong:

[E::hts_idx_push] Invalid record on sequence #2: end 19212036 < begin 33141291

I raised the issue on bcftools that said it was a Delly bug.

[tobiasrausch]

No, I just added it to the GitHub code a couple days ago. I will release a new version soon but there are some other open issues that need to be fixed first. Thanks.