The allele with index 2 is not defined in the REF/ALT columns" error after AF filtering with gnomAD Exome

Question

[erikatatianacs]

I'm running a germline variant calling pipeline with bcftools, and I annotate/filter variants based on allele frequency from gnomAD exomes (gnomad.exomes.r2.1.1.sites.vcf.bgz). After filtering, I get an error in IGV when loading the VCF:

Error loading features for interval: chr1:114940611-114940651 htsjdk.tribble.TribbleException$InternalCodecException: The allele with index 2 is not defined in the REF/ALT columns in the record

chr1 114940632 . C CA 115.328 PASS INDEL;IDV=32;...;AC=1,1;AN=2;... GT:PL 1/2:176,86,79,72,0,93 This genotype 1/2 suggests two alternate alleles, but only one ALT allele CA is listed.

Here are the key steps I use:

    # 1. Variant calling
bcftools mpileup -Ou -f genome.fa sample.bam |
bcftools call -mv -Ou |
bcftools filter -s LowQual -e 'QUAL<30 || DP<10' -Ov -o sample_raw.vcf

# 2. Compress + index
bgzip sample_raw.vcf && tabix -p vcf sample_raw.vcf.gz

# 3. Annotate with gnomAD (AF)
bcftools annotate \
  -a gnomad.exomes.r2.1.1.sites.vcf.bgz \
  -c CHROM,POS,REF,ALT,INFO/AF \
  sample_raw.vcf.gz -Oz -o sample_annotated.vcf.gz
tabix -p vcf sample_annotated.vcf.gz

# 4. Normalize (split multiallelics)
bcftools norm -m -any sample_annotated.vcf.gz -Oz -o sample_norm.vcf.gz
tabix -p vcf sample_norm.vcf.gz

# 5. Filter by AF and quality
bcftools view -f PASS -i 'INFO/AF<0.05' sample_norm.vcf.gz -Oz -o sample_filtered.vcf.gz
tabix -p vcf sample_filtered.vcf.gz

# 6. (Optionally) Normalize again?
bcftools norm -m -any sample_filtered.vcf.gz -Oz -o sample_final.vcf.gz
tabix -p vcf sample_final.vcf.gz

I ALREADY CHECK THIS: VCF headers are intact (#CHROM, INFO, FORMAT, etc.)

bcftools view doesn't complain about the file.

IGV loads unfiltered/raw VCFs without issues.

Manual checks show many sites with 1/2 genotypes but only one ALT allele.

Normalization (bcftools norm -m -any) doesn't seem to resolve the mismatch in some cases.

I dont know where is my mistake. Any advice is appreciated, especially from anyone who's successfully integrated gnomAD Exome annotations into a clinical/exome pipeline without IGV errors!

[Pierre Lindenbaum]

that would be a problem with bcftools norm. What is your version of bcftools ? Furthermore, show us the output of

bcftools view -G --no-header sample_norm.vcf.gz "chr1:114940632-114940632"

[erikatatianacs]

I am using bcftools 1.19 Using htslib 1.19

the ouput

chr1 114940632 . C CA 115.328 PASS INDEL;IDV=32;IMF=0.820513;DP=39;VDB=0.108342;SGB=-0.69312;RPBZ=-2.30802;MQBZ=0;MQSBZ=0;BQBZ=-1.75759;SCBZ=0;MQ0F=0;AC=1,1;AN=2;DP4=0,6,0,32;MQ=60;AF=0.000890251

[Pierre Lindenbaum]

there was already a problem before bcftools nom . Do you still have two alleles in sample_raw.vcf.gz at the same position ? if true you should upgrade bcftools and if the problem persists you should submit a bug report.

[Pierre Lindenbaum]

by the way, when using bcftools annotate this way, you would overwrite the original AF field with the value of gnomad.

https://samtools.github.io/bcftools/bcftools.html#annotate

The imported VCF annotations can be renamed as "DST_TAG:=SRC_TAG" or "FMT/DST_TAG:=FMT/SRC_TAG".

[erikatatianacs]

yes, but I think that I need it in this way because the main idea if filter by the population AF, so I dont need the AF of the sample