Phailed SV phasing

[agatafant]

I am using the complete wf-human-variation pipeline v.2.2.4 (previously I used dorado 0.6.0) with --sv --snp --phased --cnv options. I have 3 affected siblings (A1, A2 and A3) with both the deletion and the duplication; sample H1 has only the deletion. The deletion is of 405kb (Chr6:162154938-162560714​) and the duplication of 509kb (Chr6:162185434-162695422​). Keep into account that this 2 SVs have a big overlap. After having added in nextflow config Sniffles2 options (--long-del-length 520000 and --long-dup-length 520000), the 2 SVs are detected but, except in one case, it isn't able to phase (actually I know,from different experiment, that A1, A2 and A3 has inherited the del from the father and dup from the mother ).

Since the two SVs are in two different alleles for A1, A2 and A3, I was expecting 1|0 for one and 0|1 for the other. For A2 I obtain the right genotypes also with phasing. For the other 2 I obtained:

• for A3: 0/1 for DEL and 1|0 for DUP
• for A1: 1/1 for DEL and 1|0 for DUP

In sample H1 (that has only the deletion) it identifies the DEL with GT:1/1.

Here below IGV screenshots (color by HP and sort by HP) of beginning and end coordinates of del and dup for A1 and H1 samples.

Why does it call the deletion in homozygosis? Why isn't able to phase?

[fritzsedlazeck]

Hi ,
the phasing option in sniffles is dependent on a phased bam file. I am just saying this because I am not sure if that was the input.

To generate that you would need to call SNV and phase them (e.g. Clair + Whatshapp phasing and tagging). Then put the tagged bam file into Sniffles.

We just released a new version of sniffles with some corrections on the phasing logic. So maybe worth a try as well.
Thanks
Fritz

[fritzsedlazeck]

@agatafant can you confirm if you used the phase tag in sniffles ? In the current version that tag is required to parse out and utilize the phasing information from the bam file.
Thanks
Fritz

[agatafant]

Hi @fritzsedlazeck , I run sniffles 2.5.2 by itself and the overall situation is definitely better. Here a picture with the genotype results before and now with v2.5.2. In red if the genotype is not correct, orange if the genotype is correct but no phasing available, green if genotype and phasing are ok, in purple if I don't if is correct or not because we don't have validations.

• For samples from PDM1O till PDM50 is now almost all ok (except for deletion of PDM1N that hasn't been phased.. maybe for low coverage?).
• For samples PDM1F, PDMOV and PDMOW now the genotypes are all ok, but not phased: especially for PDM1F, that has very high coverage and good N50, sounds strange that is not able to phase.
• Also for PDM6X isn't able to phase the deletion (maybe here for bad N50?).
• For PDM7D since I have a partial overlapping of deletion and duplication sounds strange: deletion is ~116kb, duplication is ~94kb and the overlapping part is ~31kb

thank you!