Open damiansm opened 9 years ago
That would make it easier for many users. We should definitely also consider working out a really nice online tutorial. There are a number of nice templating systems, I think Manuel has used several of them?
With single vcf we will not have a combined calling. This will reduce the performace!
Of course, uploading them separately will make it more easier for lot of users. But it will reduce the hurdle to make a combined calling.
I think providing a tutorial how you can generate multivcf files from single once and how you can do a multiple calling and discuss the benefit of it will be my favorite option.
As ever I am ignorant of the steps before Exomiser.
Max - are you talking about the reduction in performance of variant calling when you use single VCFs as you can can't take advantage of the fact that a variant is seen in a parent and child to strengthen the case for it being real?
Sounds like a simple tutorial on how to use vcf_merge or other tools is a better option than implementing trio input as separate VCFs?
@damiansm no. I mean 3 variant calling on its own is different to multiple vcf calling on the complete trio at the same time. If separate variant calls are made we do not know if a mutation is ref or not covered in the parents. This increase the false positive rate.
Merging will always be vcf calls on its own and therefore not the best way.
But in reality, clinicians only have one vcf per person And do not have the ability to make a combined calling. These clinicians will be stuck and I like your idea to help these guys. Otherwise they will only upload the index and look up the mutations in the parents... much work...
Using separately called VCF files would be a reasonable option if they are gVCFs. It would be better than nothing if they are plain VCFs. We should maybe put a tutorial on how to do things "right" but also offer a quick solution with a warning...?
Can Exomiser handle a 3 generation family (daughter, mom, and grandma)? My current understanding of how a trio would be run is creating a ped file, merging the vcfs together, and only using the proband's HPO IDs. Is that correct?
I guess I do not quite fully understand what issues would arise from merging the vcfs together. You would still be retaining each sample's variant calls and associated data (http://samtools.github.io/hts-specs/VCFv4.1.pdf), so I'm not quite sure why this would affect performance outside of a convenience standpoint.
Has there been any update on this? Allowing vcfs to remain unmerged would be beneficial.
Should we allow the ability to upload patient and parent exomes as 3 separate VCF files rather than a multi-sample. Or is it easy enough for people to use vcf_merge?