Estimating contamination - high depth targeted panel

[Meghana V]

Hi,

   I am running verifybamid on high depth targeted panel data. I have around 1500 markers on the omni vcf file  and I am setting the maxDepth to 1500.

   I have 2 issues I am encountering:

  

   1. We identified some samples with high contamination (20%). When I downsample the bam files - the estimate comes down (<0.04%).

   2. Many samples have a freemix estimate of 0.04999 again when this data is downsampled to use 50% of the data, the estimate seems reasonable.

  

Here is my command:

verifyBamID --vcf $target.intersect.vcf --bam $bam --maxDepth 1500 --site --precise --best --out $sample.contamination

Please let me know if there is something I am missing (or) could this be an issue with maxDepth parameter (or)  artifacts that arises from high depth data?

Happy to provide more information.

Thanks,

Meg