VCF multiple records with the same POS, why?

[krawitz]

Hi,
I stumbled upon two entries in the VCF of NA12878 with the same position. I thought this is not allowed. I checked the vcf specifications at http://www.1000genomes.org/wiki/Anal...mat-version-41 and it actually says:
"It is permitted to have multiple records with the same POS".
Does anyone know what the advantage is of allowing mutliple records with the same POS? One could easily merge multiple records with the same POS into a single one. Mutliple alleles would then be listed in the ALT column and different genotypes in multiple vcf files would then point to these different alleles (0/1, 0/2, ...). Actually that is what vcf-merge is already doing now.
Am I missing anything?
Thanks a lot!
peter

[krawitz]

example

the example of NA12878 with the two POS is:

10 46999406 . A G . . . 1/1
10 46999406 . AGGG AGGGAGGTGGGGG 0/1

In my opinion one of the called genotypes must be an artifact, as this is an diploid individuum and can therefore only have one genotype at this position.

what do you think?

[BAMseek]

I have run into something like this, too. An insertion occurs between two adjacent positions. I think VCF lists the position before the insertion. So for example, if there was a substitution followed by an insertion, they would get the same position. You could have a situation where there is a homozygous substitution followed by an indel, and they would all be labeled as the same position, although it wouldn't make sense to combine them. Maybe it is easier to think of the insertion as position "46999406.5" (in your case, looks like it actually might be 46999409.5).

Justin