Reference allele is too long message GATK

Question

[win]

hi all, i generated a gVCF file using HaplotypeCaller. When i run ValidateVariants from GATK i get the message

"Reference allele is too long (108) at position chr9:99423855; skipping that record. Set --reference_window_stop >= 108"

Any ideas what is causing this?

Command used to generate the gVCF

java -Xmx16g -jar algorithms/gatk3/gatk3.8.jar -T HaplotypeCaller -R references/hg38gatkbundle/Homo_sapiens_assembly38.fasta -I data/HG100/HG100.output.bam --emitRefConfidence GVCF --dbsnp references/hg38gatkbundle/Homo_sapiens_assembly38.dbsnp138.vcf -o data/HG100/HG100.output.raw.snps.indels.g.vcf --reference_window_stop 1000

Command used for validation

java -jar algorithms/gatk3/gatk3.8.jar -T ValidateVariants -R references/hg38gatkbundle/Homo_sapiens_assembly38.fasta -V data/HG100/HG100.output.raw.snps.indels.g.vcf --dbsnp references/hg38gatkndle/Homo_sapiens_assembly38.dbsnp138.vcf --validationTypeToExclude ALLELES

Any help highly appreciated.

Answer 1

[WouterDeCoster]

My guess is a long deletion, with 108 nucleotides in your reference allele field. As suggested by the error message setting --reference_window_stop to a value >= 108 should work.

[win]

OK, but there are other such messages as well, same message but differing length like some say 108, some are 150 etc. I have also set --reference_window_stop to 1000.

[WouterDeCoster]

And does that help for these errors?

[win]

No, thats the problem.

[WouterDeCoster]

Then, what error messages do you get with --reference_window_stop 1000?

[win]

same message, it seems like the --reference_window_stop is not working or has incorrect values.

[Satyajeet Khare]

This worked fine for me.

java -jar  GenomeAnalysisTK.jar -T ValidateVariants --reference_window_stop >= 300 -R Genome.fa --variant:VCF All.vcf.gz