GATK 3.8.0 VariantFiltration ERROR input string

Question

[rturba]

Hello,

I am running GATK 3.8 on a Linux cluster. I am trying to run the tool VariantFiltration with the following commands:

java -jar $GATK/GenomeAnalysisTK.jar \
    -T VariantFiltration \
    -R $ref \
    -V $wd/vcf/filtered/stick84_SNPs_GATK.vcf \
    --filterExpression "QD < 2.0" --filterName QD2 \
    --filterExpression "FS > 60.0" --filterName FS60 \
    --filterExpression "MQ < 40.0" --filterName MQ40 \
    --filterExpression "MQRankSum < -12.5" --filterName MQRS-12.5 \
    --filterExpression "ReadPosRankSum < -8.0" --filterName RPRS-8 \
    -o $wd/vcf/filtered/stick84_SNPs_filtered_GATK.vcf

However, it is throwing me this error:

ERROR MESSAGE: For input string: "nan"

I don't know what this error means and how to fix it. All the potential help I find online sends me to ghost pages of GATK that I cannot access :(

Please, any help will be greatly appreciated! Thank you and take care!

Answer 1

[rturba]

Ah, I've found out the reason for that here: https://github.com/broadinstitute/gatk/issues/5582. I hope this helps others as lost as me! I did not realize the INFO fields could have "NaN" values. Solution to fix is:

bcftools view in.vcf.gz |
sed 's/=nan/=NaN/g'  |
bgzip > fixed.vcf.gz

[GenoMax]

I moved this to answer so you can accept it (green check mark) to provide closure to the thread.