VQSR with extremely low TPs (<0.01), How low can tranche goes?

Question

[greyman]

Hi, Sorry that i have posted this at here because for some unknown reason i am unable post a question in GATK forum. Any help from here is truly appreciated.

I have 45108 variants from 31 exome vcf files. After VariantRecalibrator, my tranche specificity shows the best Ti/Tv ratio at 57.8 (with Ti/Tv ratio of 2) which is very different from several examples from the tutorial with tranche value 90 - 99.9. Is it normal and should i proceed with ApplyVQSR with -tranche cutoff level at 57.8? Is the value acceptable?

gatk --java-options "-Xmx48g -Xms48g" VariantRecalibrator -R $reference_dir -V tmp_sitesonly.vcf.gz -O tmp_sitesonly_recal_snps.vcf.gz --tranches-file tmp_sitesonly.snps.tranches -an MQ -an DP -an QD -an FS -an ReadPosRankSum --max-gaussians 4 --trust-all-polymorphic -tranche 60 -tranche 55.9 -tranche 55.8 -tranche 55.7  -tranche 55.5 -tranche 55 -mode SNP --rscript-file tmp_Merged_snps.plot.R --resource hapmap,known=false,training=true,truth=true,prior=15.0:$hapmap --resource omni,known=false,training=true,truth=false,prior=12.0:$omni --resource 1000G,known=false,training=true,truth=false,prior=10.0:$oneksnp --resource dbsnp,known=true,training=false,truth=false,prior=2.0:$dbsnp