Too many Variants called by HaplotypeCaller GATK

[drmaly]

Hi everybody,

I am doing Exome analysis on 8 individual (family) and I just got something weird as I think. My pipeline is following the best practices of GATK but when I checked my last vcf file after calling with HaplotypeCaller walker, I found the file contains 2.1 million lines (am assuming these are all variants) . My vcf file is a merged file for all of the 8 samples together. As I knew Exome pipeline should usually get you around 20 thousand variants so am really confused now on what is wrong in my pipeline. Any ideas will be really appreciated.

[vdauwera]

when I checked my last vcf file after calling with HaplotypeCaller walker, I found the file contains 2.1 million lines

Hi there,

If you consider the HaplotypeCaller output to be your final file, you're not really following Best Practices... As indicated in the GATK documentation, the output of HaplotypeCaller is a set of raw variants that is likely to include a lot of false positives. You have to filter them (either with VQSR or hard filters) to generate a callset with the desired level of sensitivity and specificity.

[jp.]

vdauwera is right
VCF output might differ if your commands are not proper from the begining (BWA to HaplotypeCaller to V). could you post your command and message you got before finishing the each job to help understanding what happened there

jp.

Hi everybody,

I am doing Exome analysis on 8 individual (family) and I just got something weird as I think. My pipeline is following the best practices of GATK but when I checked my last vcf file after calling with HaplotypeCaller walker, I found the file contains 2.1 million lines (am assuming these are all variants) . My vcf file is a merged file for all of the 8 samples together. As I knew Exome pipeline should usually get you around 20 thousand variants so am really confused now on what is wrong in my pipeline. Any ideas will be really appreciated.