Handling other genotyping formats?

[smlmbrt]

The current pipeline accepts the following types of genotyping data as input:

• PLINK 1 binary (.bed/bim/fam) : https://www.cog-genomics.org/plink/2.0/input#bed
• PLINK 2 binary (.pgen/pvar/psam) : https://www.cog-genomics.org/plink/2.0/input#pgen
• Variant call format (.vcf) : assuming it's from imputed genotyping data (e.g. the output of Michigan/TopMed imputation server, etc.)

We're wondering if these options are sufficient, or if we're missing other useful inputs? Other possibilities could include:

• BGEN format : https://www.well.ox.ac.uk/~gav/bgen_format/index.html [more compressed than plink]
• Genomic VCF (gVCF) : https://gatk.broadinstitute.org/hc/en-us/articles/360035531812-GVCF-Genomic-Variant-Call-Format [carries information about regions that have been called but are not listed as variant sites/lines (e.g. homozygous reference) which would be necessary for tabulating variant matching statistics] Discussed previously here: format for input data includes reference calls? #50

[moe1619]

one vote for gVCF. Helpful for calculating PRS when starting from WGS for example. Thank you!

[bgulko]

This is my workflow as well (WGS->gvcf->PGRS). Building a VCF by filtering our gVCF to induce a VCF call from every position mentioned in any PGR catalog model is definitely troublesome. Second vote for gVCF.

[bnwolford]

Vote for bgen! As an alternative, I could use a wrapper that creates temporary PLINK files from bgen for just the markers needed for the PGS(s) requested.