This site is now read-only. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!
--maxGaussians parameter in GATK VQSR
I have been running VQSR on a very large dataset (160M variants from 12K WGS). I am having problem in understanding the output of VQSR. I followed the exact parameter settings mentioned here (except --sample-every-Nth-variant 10) -
I am getting quite different results (which I don't expect) by using different Gaussians. See the comparison plot below -
The default value (i.e., 8) produces similar plot like --maxGaussians 6. Do you have any insight on this ? Don't I have enough variants for use more Gaussian values (for more clusters) ? Why these results are so different?
More surprisingly, if I randomly remove only 20 sites out of 160M, the plot looks different (both with same Gaussian value - 6). See below -
My question is - is it expected ? Can only 20 variant difference in 160M variant file create so different results. My input files have exactly same variants (one with less 20 variants) with same annotations. Any help will be highly appreciated.
Thank you very much.