I compared HaplotypeCaller and UnifiedGenotype of the latest version of GATK (2.7-4) on the same NA12878 trio. It seems HC missed some true variants (specifically, 7 out of the 49 experimentally validated de novo SNVs are missed). I guess this might be partly due to HCMappingQualityFilter, because the log says ~9% of the total reads have been filtered by this filter, and it looks there is not such a filter in UnifiedGenotyper. I wonder is there a safe way to relax HCMappingQualityFilter (default 20) to achieve a better sensitivity for HC?
Thanks in advance!