Let's say 50 samples were sequenced at 25x. After a while, the group realizes that the experiment is under-powered and adds 100 samples more, but at a coverage of 5x. I'm talking about WGS.
The scientist runs the samples following GATK's best practices until he/she is done running haplotypeCaller on them.
Then the question is how to proceed:
a) combine all 150 gVCFs and do joint calling. Then do site filtering followed by genotype filtering.
b) Do joint calling with the high coverage samples and in parallele do joint calling with the low coverage samples. Then do site filtering, merge both VCFs and filter by genotype.
c) combine all 150 gVCFs and do joint calling. Split VCF into two according to coverage and do site filtering. Merge both VCFs and filter by genotype.
Option "a" sticks to GATK's recommendations, but it ignores the high difference in coverage between sample sets. The opposite is the case for option "b". And option "c" is like a hybrid approach of "a" and "b".
I guess what I want to ask if it is an issue to do joint-calling and then site-filtering when groups of samples have large differences in coverage.