Use of GenotypeGVCFs in population genetic studies

Question

[kk.mahsa]

I have 16 whole genome sequenced samples from two populations (8 for each population). My goal is detection of signature of selection and introgression. I performed read cleaning, mapping to reference, mark duplication. SNP calling was performed using HaplotypeCaller in GATK for each sample separately. Now my question: For downstream analysis (PCA, ADMIXTURE analysis and detecting signature of selection), do I need to use GenotypeGVCFs command in GATK for genotype joining? Or I can create one VCF file per sample separately (without GenotypeGVCFs) and merge them for downstream analysis after variant filtering?

Thanks in advance

[tothepoint]

Use GenotypeGVCFs file for post analysis.

The key difference between a regular VCF and a gVCF is that the gVCF has records for all sites, whether there is a variant call there or not. The records in a gVCF include an accurate estimation of how confident we are in the determination that the sites are homozygous-reference or not.

[kk.mahsa]

Thanks tothepoint, Now, if i produce separate VCF file for each sample, how can i merge them? merging must be based on population?

[tothepoint]

Merge or combine vcf file? GATK combinegvcfs will do the job for you.

[vdauwera]

You should produce a gVCF for each sample (using haplotypecaller in GVCF mode) then (EDITED) combine them in order to run GenotypeGVCFs on all of them together.

(Edited to correct a mistake)

[tothepoint]

vdauwera please correct me if I am wrong. We can run GenotypeGVCFs after CombiningGVCFs. I did perform some analysis following the explanation in GATK page describing:

The GATK4 GenotypeGVCFs tool can take only one input track. Options are 1) a single single-sample GVCF 2) a single multi-sample GVCF created by CombineGVCFs or 3) a GenomicsDB workspace created by GenomicsDBImport. A sample-level GVCF is produced by HaplotypeCaller with the -ERC GVCF setting.

[vdauwera]

Oh I misread that as CombineVCFs (without the G), sorry. Yes you’re correct. I would recommend using the GenomicsDB method (that’s what I had in mind, realizing now I didn’t write it out — need coffee) rather than the basic combiner tool, but both are valid.

I edited my previous post to minimize confusion if someone else sees this thread.