Splitting VCF by sample: different files but same SNPs number

Question

[greed]

Hi there. I'm working with VCF files and I've noticed a peculiarity: using the following commands: bcftools view -S sample_A_list.txt input.vcf > sample_A.vcf and bcftools view -S sample_B_list.txt input.vcf > sample_B.vcf, I've created these 2 VCF files deriving from input.vcf, splitting it into 2 different samples (A and B). sample_A.vcf has 27 samples and sample_B.vcf has 116 samples. Now if I run egrep -v "^#"sample_A.vcf | wc -l or egrep -v "^#"sample_B.vcf | wc -l in order to have the number of SNPs for each VCF, I collect the same result: 5997 SNPs for both files. Then I pruned the SNPs in linkage using the plink pipeline (R 0.4) in order to get 2 new VCF via the recode function. Running the same bash command to get the total number of SNPs post pruning I get values that are totally different: sample_B.vcf (with 116 samples) post LD-pruning passes from 5997 SNPs to something more than 2000, while sample_A.vcf (with 27 samples) post LD-pruning passes from 5997 SNPs to something more than 500. So the first question is: if I have 2 different VCFs, for which reason I've got the same SNPs total number? The second questions is: for what reason I have this huge difference in SNPs number post LD-pruning between my 2 files? Thank you for the answers and the help.

Answer 1

[Pierre Lindenbaum]

if I have 2 different VCFs, for which reason I've got the same SNPs total number?

because bcftools view --sample SAMPLEX will not remove the variant if the only genotype (for SAMPLEX ) is homozygous on REF (0/0). You need an extra step like

bcftools view -i 'AC>0' one_sample.vcf

[greed]

Ok, so if I correctly understand I have to do this step for each VCF (sample_A.vcf and sample_B.vcf) generated from view function.