Combining inhouse data with 1000 genome for PCA

Question

[ankita]

For PCA analysis, I have common coordinates between Inhouse exome data and 1000 Genomes data (Phase 1). I want to retrieve genotypes for those common SNPs from VCF files in 1000 genomes [then convert to plink for smartPCA). I thought of a option of converting all VCF files in phase 1 to ped which is very memory intensive. What can be the possible solution for this problem?

[ankita]

thanks, I will try this :)

[Giovanni M Dall'Olio]

Consider using the 1000Genomes data for imputing the genotype of the SNPs missing in your dataset: http://www.1000genomes.org/faq/can-i-use-1000-genomes-data-imputation

Answer 1

[chrchang523]

PLINK 1.9 supports direct conversion of VCF to .bed+.bim+.fam , which should be readable by smartpca. For example,

plink --vcf ALL.chr1.phase1_release_v3.20101123.snps_indels_svs.genotypes.vcf.gz --out 1000g_chr1

Answer 2

[ankita]

Hi

I am using vcf-concat to combine all VCF files from 1000 Genomes and then using plink 1.9. Is vcf-concat option is the only way for concatenation, is there any other way in plink 1.9 because it is much faster ?

[chrchang523]

You can use plink --merge-list.

[ankita]

I have used --merge-list but it throws an error and warnings for snp inconsistencies. What I understand is merging is used when 2 files have same snps and you have to merge data for different individuals with same snps while I want just want to concatenate two vcf files from different chromosomes say chr1 and chr2 for making one single plink file, in short I want to concatenate all vcf files of 1000 g to make one plink file.

[chrchang523]

bcftools concat may be your best bet. (--merge-list also handles concatenation, but each variant needs to have a unique ID.)