How to generate "dummy"/"fake" VCFs for testing software?

Question

[kynnjo]

I need to generate "dummy"/"fake", but "formally valid", VCF data to test the performance of processing pipelines.

The need for such data arises in many context, but at the moment I am most interested in measuring the performance of alternative approaches to merging large numbers (>10K) of single-sample VCFs.

Most of the VCF data that I have ready access to is protected patient data, which limits what I can do with it (e.g. which cloud servers I can upload it to for processing).

Can anyone recommend a method for generating dummy/fake single-sample VCFs?

[4galaxy77]

1000 genomes?

Answer 1

[Pierre Lindenbaum]

https://gnomad.broadinstitute.org/downloads in gnomad v3

HGDP + 1KG callset. These files contain individual genotypes for all samples in the HGDP and 1KG subsets.