long view SNPs in vcf file

Question

[young_bioinformatician]

Hello guys,

I did run freebayes and performed some filtration on the output. - removing indels, and discardmulti-allelic sites. There were long-looking SNP sequences with multiple allelic sites before filtration. For example, ATTGC, ATGGC, TTGGC..

However, even after discarding multiallelic sites, the appearance of vcf file is still same that you can see the picture below. How can I handle it? I will convert my vcf to plink and implement PCA. Would it be problem with those sites?

Also, my another question regarding discarding multi-allelic sites. How is it decided which site is discarded or not ? Is there any way that I should follow and consider the sites when discarding them ? Or can I still go with multi-allelic site for PCA analysis?

Thanks!

[Pierre Lindenbaum]

However, even after discarding multiallelic sites,

what was the command line ?

How is it decided which site is discarded or not ?

it depends of the scientific question asked

[young_bioinformatician]

Thank you for your answer, Pierre.

Not the question, but I have metagenomic data. So multi-allelic sites could make sense in my case. However, I have an error about the multi-allelic when I merge two vcf file to create PLINK format. Therefore, I always discard these sites from VCF, too.

In addition to this, can I directly implement PCA on data which contains multi-allelic sites?

Answer 1

[chrchang523]

You are misunderstanding what "multiallelic" means.

"Multiallelic" does not mean "an allele has multiple bases", which occurs at positions 358 and 399 in your screenshot. It means "there are multiple ALT alleles", e.g. REF is "A" while ALT is "G,T". If there is no comma in the ALT column, the variant is not multiallelic.

As for why the bcftools filters you applied did not remove the variants at positions 358 and 399, that's because those variants are actually regular SNPs, they're just represented weirdly; see https://genome.sph.umich.edu/wiki/Variant_Normalization for some discussion. You can use the "bcftools norm" command with the --fasta-ref flag to left-normalize these weirdly-represented SNPs.

[young_bioinformatician]

Thank you. bcftools norm --fasta-ref was exactly what I was looking for, It worked well !

Sorry, I know what multi-allelic means but I think I could not explain my issue very well in the question post. The picture I shared was obtained after multi-allelic sites were discarded.

[yzliu01]

I got the sample problem. Very helpful!