Generation of incorrect heterozygous calls after left normalization using bcftools

Question

[nkausthu]

I have following records in one of the gvcf file

1       3753032 .       GTTTT   G,GT,GTTT,GTTTTT,GTTTTTT,GTTTTTTTT,<NON_REF>

1       10502954        .       CTTTTT  C,CT,CTTT,CTTTT,CTTTTTT,<NON_REF>

1       11272829        .       T       <NON_REF>

1       11272839        .       G       <NON_REF>

1       15978128        .       T       <NON_REF>

1       15978129        .       T       <NON_REF>

1       38332078        .       T    TCA,TCTCA,TCACACACACACACACACA,TC,<NON_REF>

1       67725648        .       GAAAA   G,GAA,GAAA,GAAAAA,GAAAAAA,<NON_REF>

1       72748277        .       ATT     A,AT,ATTT,ATTTT,ATTTTT,ATTTTTT,<NON_REF>

1       150782110       .       CAAAAA  C,CA,CAA,CAAA,CAAAA,<NON_REF>

1       155724315       .       GTT     G,GT,TTT,GTTTTT,GTTTTTT,<NON_REF>

1       158058266       .       CTTTTTT C,CT,CTT,CTTT,CTTTT,CTTTTT,<NON_REF>

1       201082902       .       C       CAA,CAAA,<NON_REF>

1       212618993       .       A       C,<NON_REF>

1       237955682       .       C       CGTGT,CGTGTGT,<NON_REF>

2       27532239        .       CAAA    C,CA,CAA,CAAAAAAAAAAAAAAAAA,<NON_REF>

2       47641559        .       TAAAAAA T,TA,TAA,TAAA,TAAAA,TAAAAA,<NON_REF>

2       100058714       .       CAA     C,CA,CAAA,CAAAA,CAAAAAAA,CAAAAAAAA,<NON_REF>

2       113303450       .       T       <NON_REF>

2       113303451       .       G       <NON_REF>

2       207998878       .       AT      A,ATT,ATTT,ATTTT,ATTTTT,ATTTTTT,<NON_REF>

2       231333532       .       CAAA    C,CAA,CAAAAAAA,<NON_REF>

3       42734487        .       G       <NON_REF>

3       42734750        .       C       A,<NON_REF>

3       42734751        .       C       <NON_REF>

3       47484723        .       TACACACAC       T,TAC,<NON_REF>

when we have done left normalization using bcftools after joint genotyping, lots of false heterozygous calls has been generated with no reads supporting the altered allele as follows

0/1:14,0:42:72:149,0,164

0/1:10,0:35:17:88,0,111

I guess it's due to incorrect splitting of multi-alleles. It would be great if anyone can suggest ways to remove these variants from downstream vcf file ?

[finswimmer]

Hello,

please provide an example dataset one can use directly for testing.

Thanks!

fin swimmer

[nkausthu]

I can provide vcf file after left normalization is that sufficient?

[finswimmer]

Hello,

that's better then nothing. But the input vcf would be more useful. Reduce it to some example lines that show your problem.

Answer 1

[Kevin Blighe]

Hey,

With multi-allelic sites, I think that it is better to do this in a 2 step process, and you must also make use of the reference genome against which the variants were originally called.

So, something like this:

# 1st pipe, splits multi-allelic calls into separate variant calls
# 2nd pipe, left-aligns indels and issues warnings when the REF base in your VCF does not match the base in the supplied FASTA reference genome
bcftools norm -m-any myvariants.vcf | \
  bcftools norm -Ob --check-ref w -f /ReferenceMaterial/1000Genomes/human_g1k_v37.fasta \
  > myvariants_norm.bcf ;

Kevin