Filtering VCF files based on VAF giving incorrect results

Question

[Arton]

I'm trying to filter variants that have VAF<0.2. I used picard FilterVcf, bcftools filter, and tried to use GATK VariantFiltration, and neither of these worked. picard and bcftools only filter a portion of variants but I still get a long list of variants with VAF<0.2 labeled as "PASS". I haven't figured out how to use GATK VariantFiltration for VAF. Any help to figure out the problem would be greatly appreciated. Here are my codes:

java -jar picard.jar FilterVcf  \
    I=input.vcf
    MIN_AB=0.2 \
    O=output.vcf   

bcftools filter --include '(FMT/AD[0:1])/(FMT/AD[0:0]+FMT/AD[0:1]) >=0.25 && (FMT/DP[0:0] >=10)' \
        input.vcf \
        --output output.vcf \
        --soft-filter Filter1 \
        --mode +

Answer 1

[Pierre Lindenbaum]

using vcfilterjdk: https://jvarkit.readthedocs.io/en/latest/VcfFilterJdk/

 java -jar ${JVARKIT_DIST}/jvarkit.jar vcffilterjdk -e 'return !variant.getGenotypes().stream().filter(G->G.isHet() && G.hasAD() && G.getAD().length==2).anyMatch(G->{double n1=G.getAD()[0];double n2=G.getAD()[1];if(n1+n2<=0) return false; double f=(n2/(n1+n2));return (f<0.2 || f>0.8);});'  --filter Filter1 input.vcf

[Arton]

Thank you. I can't try a completely new tool but I will keep your command for future use. Do you think my commands were correct? For example what does mean that MIN_AB is double?

[Pierre Lindenbaum]

Do you think my commands were correct?

I'm not familiar enough with this bcftools syntax.

but what if there is no AD, what if both are '0', what if there is a NO-CALL at this position ? what if the variant is multiallelic ?