Request: multisample vcf and existing alignement

[leone93]

Hello, thanks for the tool!
However, I have two ideas that could be nice to implement in the software.
The first is related to the use of multisample vcf. I think will be interesting if it can work with multisample vcf (for instance, the one produced by survivor) to have a clear representation of the genotype for each variant for each sample. And also discriminate where the sample homozygous and where the is no information about the genotype. This will also be nice using the capability of multi alignment of minigraph.
The second point is more related to old version of SVjedi; I think will be nice to be able to use other alignments (maybe in bam format) instead of doing a new one. I mean, to obtain the VCF, you probably have the alignments where it comes from, so it will be nice to use these alignments to genotype the SVs. Thinking about multisample will be nice to give the software a list of bam, the same present in the vcf and genotype the multisample.
Let me know what you think,
Thanks
Leo

[clemaitre]

Hi Leo,

thank you for your comment and ideas and sorry for the late reply.

Concerning the multisample idea, SVJedi-graph can already take as input a multisample VCF. SVJedi-graph uses only the description of the SVs in the VCF file, it does not use the genotype column(s). For the moment it estimates the genotype of each SV for only one sample given as one or several sequencing read files (fastq). If you want to get the genotypes for several samples, the only way for the moment, is to run SVJedi-graph several times (one for each sample) and then combine the genotype columns of all output VCF in a single VCF file. We are thinking of automating this process in a single SVJedi-graph run, but we have not found the time to dot it yet...

Concerning the second point, this is not possible to re-use a previous bam file obtained by mapping the reads to the reference genome, because SVjedi does not map reads to the reference genome but to a specific set of sequences obtained from the SV descriptions. It maps simultaneously reads on the reference and alternative allele sequences, to avoid any biais towards the reference alleles. Re-using a previous bam with reads mapped only to reference alleles would introduce such a bias.
By the way, we recommend to use SVJedi-graph rather than SVJedi because it obtains much better results, especially when there are close or overlapping SVs in the input SV call set.

Best,
Claire