Quality Control of VCFs that used different genotyping arrays

Question

[Shane]

I have three VCFs. Two of these VCFs were generated using the Precision Medicine Research Array (PMRA) and refer to SNPs as AX numbers. I was able to merge the two PMRA VCFs together.

Merged PMRA VCFs (Total genotyping rate is 0.924427):

1   AX-150343089    0   837711  T   C
1   AX-149471710    0   837756  T   G
1   AX-40234919 0   844647  G   T
1   AX-114086366    0   846320  A   G

However my Global Screening Array (GSA) VCF uses rs ids:

1   rs6605059   0   984039  T   C
1   rs4970414   0   984121  G   T
1   rs116781904 0   984475  A   G
1   GSA-rs61770779  0   984547  A   G

I was able to merge using bcftools merge pmra.vcf.gz gsa.vcf.gz -o gsa_pmra.vcf --force-samples

gsa_pmra.vcf (Total genotyping rate is 0.534432)

1   AX-29796323 0   1117607 C   G
1   GSA-rs61766344  0   1118711 T   C
1   AX-29797251 0   1120162 A   G
1   AX-29797373 0   1120521 A   C
1   AX-38925889;rs9442373   0   1127258 C   A
1   AX-29801021;rs4072537   0   1129916 T   C
1   AX-29801231;GSA-rs11260598  0   1130346 C   T
1   AX-29801717 0   1131207 T   C
1   AX-107792172    0   1133289 GT  G
1   AX-107792172    0   1133290 G   TT
1   rs61766346  0   1133503 A   G
1   AX-29803231 0   1134155 A   G

I have 4681 samples using PMRA and 40 samples using GSA

Total Positions: 944562 (PMRA), 692246 (GSA)

• PMRA Only: 805321
• GSA only: 553005
• Overlap: 139241

My issue is that the genotyping rate (0.534432) is low and I am afraid that when I do any sort of quality control filtering, it will remove too many SNPs/samples. Does anyone have any advice/comments?

[bk11]

Are your all genotyping data in the same build? The SNPs ids with AX numbers or rsid is not a problem here. You can replace those with chr:position and later annotate into rsid. I would recommend you to use Plink to quality control before you merge your datasets. You can follow the following steps provided that your two datasets are in the same build-

1. Convert your vcf files into plink binary files-

   plink --vcf pmra.vcf.gz --make-bed --out pmra

   plink --vcf gsa.vcf.gz --make-bed --out gsa

2. Quality control your gsa.bed/bim/fam & pmra.bed/bim/fam files: I would subset data to chromosomes 1-23, get rid of AT,CG GC and TA SNPs and swap AX-number and rsids with chrom:position

3. Find common SNP and Merged your QCed data: I would find common SNPs between your gsa and pmra data, and merge them. While merging you should make sure that each alleles match for the SNPs.

[Shane]

Thanks for your reply. Yes the genotyping data is in the same build.

Trying to understand this: "get rid of AT,CG GC and TA SNPs". "While merging you should make sure that each alleles match for the SNPs", for this, do you mean that I should ensure that the ref and alt allele from GSA and PMRA SNPs match?