I am trying to compare variant calling outputs of GATK's HaplotypeCaller and DeepVariant. Their raw output is very different; for example, in a WGS sample, DeepVariant called 947386 variants located on chr1, whilst HaplotypeCaller reported only 345294 variants with the same input bam file. These raw calls include low quality variant ("False positives"); I am, therefore, curious to see whether the number is high quality calls is more alike. However, these softwares seem to output very different quality metrics for their calls. I was unable to find documentation on these metrics.
For example, one variant that was called by both callers has these quality scores:
- DeepVariant: 12.5
- HaplotypeCaller: 405.64
Could anyone point me to the proper documentation or give me a rough reference of which cutoff may be used for "high quality" variant calls for either softwares' outputs?
Thanks for your help.