Using gnomad HGDP + 1KG datasets as reference

[asgarhussain]

Do we need to make the changes in the HGDP + 1KG gnomad datasets (vcf) to usable for the ntroot ?

[lcoombe]

Hi @asgarhussain,

Thanks for your interest in ntRoot!

Could you give us a bit more information about the VCF file that you are asking about? I assume the VCF is multi-sample?

Lauren

[asgarhussain]

Hi Lauren,

I removed the samples information from the vcf file (https://gnomad.broadinstitute.org/downloads#v3-hgdp-1kg). I also checked the ntRootAncestryPredictor.pl and observed that it is using the "/_AF/" as regular expression to match the different population frequencies. gnomad vcfs do not use this format, their pattern is something like this : "AF_papuan="

[lcoombe]

Hi @asgarhussain,

Just so I make sure that am understanding correctly - are you hoping to predict ancestry from an input VCF file (--custom_vcf), or are you hoping to input a different integrated variant call set file (-l), and predict ancestry using WGS/genome/etc data?

[asgarhussain]

Yes, I want to use the different integrated variant call set file (-l) using the WGS data. I downloaded this file from HGDP + 1KG

[lcoombe]

Ok great - thanks for that context!

Here's a few lines of the integrated variant call set file that we provide:

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO
10      11501   .       C       A       .       PASS    AC=286;AN=5096;DP=20462;AF=0.06;EAS_AF=0;EUR_AF=0.06;AFR_AF=0.08;AMR_AF=0.03;SAS_AF=0.1;VT=SNP;NS=2548
10      36097   .       G       A       .       PASS    AC=992;AN=5096;DP=18607;AF=0.19;EAS_AF=0.26;EUR_AF=0.11;AFR_AF=0.18;AMR_AF=0.18;SAS_AF=0.25;VT=SNP;NS=2548
10      47049   .       GGA     G       .       PASS    AC=8;AN=5096;DP=346770;AF=0;EAS_AF=0;EUR_AF=0;AFR_AF=0;AMR_AF=0.01;SAS_AF=0;EX_TARGET;VT=INDEL;NS=2548
10      47317   .       G       A       .       PASS    AC=27;AN=5096;DP=479420;AF=0.01;EAS_AF=0;EUR_AF=0;AFR_AF=0.02;AMR_AF=0;SAS_AF=0;EX_TARGET;VT=SNP;NS=2548

So the ancestry-specific allele frequencies need to be structured <ancestry code>_AF=<frequency>, in a list separated by semi-colons

[asgarhussain]

Thanks, I will let you know. Can we add a vcf format converter ?

[lcoombe]

Hope it works OK for you!

We are certainly open to that possibility - but the issue I can see is that the format appears to vary across different VCF files from different sources? There would need to be an established convention to provide a specific converter (or adapt ntRoot to accept files in a different convention).