Hi everyone,

I am currently working with WGS data (Ultima Genomics, ~25-30x depth) for a population genetics study. I received individual single-sample VCF files for each participant.

The issue arises when I merge these individual VCFs using bcftools merge. Because these VCFs only record variant sites for each sample, the merged multi-sample VCF ends up with a massive amount of missing data (./.). When I apply a standard QC filter (e.g., --geno 0.05 in PLINK), I lose over 90% of my variants, leaving only about 900,000 SNPs out of 15 million.

I've seen some suggestions online about converting these "missing" genotypes (./.) into "reference" genotypes (0/0).

My questions are:

Is it scientifically acceptable to force-convert ./. to 0/0 in this context, given that the depth is 25-30x?

Are there specific concerns regarding Ultima Genomics' error profile (like homopolymer issues) when doing this?

Is there a better way to handle this "sparse VCF" problem if I don't have access to the original gVCFs for joint genotyping?

I would appreciate any insights or best practices from those who have handled similar WGS datasets.

Thank you in advance!