Is this an appropriate command for estimating contaminatin in a germline WGS sample using GATK?

Question

[curious]

I have 30x WGS human samples and want to estimate contamination using the GATK tools to crosscheck another software that I am using to estimate contamination, this is what I have done:

gatk GetPileupSummaries \
  -I my_sample.cram \
  -V small_exac_common_3.hg38.vcf.gz \
  -L small_exac_common_3.hg38.vcf.gz \
  -O my_sample.getpileups.table \
  --reference hs38DH.fa

gatk CalculateContamination \
  -I my_sample.getpileups.table \
  -O my_sample.contamination.table

where small_exac_common_3.hg38.vcf.gz was gotten from gs://gatk-best-practices/somatic-hg38/small_exac_common_3.hg38.vcf.gz

My main questions are:

 1. Is this an appropriate contam estimation workflow for germline human WGS?
 2. If so, is small_exac_common_3.hg38.vcf.gz a reasonable file to  pass for the -V and -L arguments?
 3. from my_sample.contamination.table I get columns "sample", "contamination", "error". If the value of "contamination" is 0.01, does this mean the sample is contaminated 1%?

Answer 1

[Kevin Blighe]

Yes, this is an appropriate workflow for estimating contamination in germline human WGS (it's the standard GATK method, though originally somatic-focused; it works well for diploid samples by modeling allelic fraction deviations at common SNPs).

Yes, small_exac_common_3.hg38.vcf.gz is the exact resource recommended by GATK for -V and -L (it's a subset of ~10k high-confidence ExAC common variants optimized for this).

And yes, a contamination value of 0.01 indicates ~1% contamination (it's the estimated fraction of reads from cross-sample admixture). For cross-checking, I'd suggest also running VerifyBamID (non-GATK alternative) if you want an orthogonal method.

Kind regards,

Kevin