Currently I am following GATK best practice for using HC 3.0+, however I'm splitting my calls to chromosomal regions (-L). Next are the following step I perform working up to GenotypeGVCF and my question.
1 - I use CatVariants (following HC) to merge all 25 chromosome gvcf files into a single gvcf file per individual.
2 - I use CombineGVCF to merge 2 .. n number of individuals together. This is done because some analysis have 300+ individuals. 3- I then use CombineGVCF again to merge all the file from step 2 into one large gvcf file for one large joint GenotypeGVCF step. 4 - GenotypeGVCF is done again based on chromosomal regions (-L), which is followed by a additional CatVariants before VQSR.
The question I have this this: Given the size of the analysis I have noticed that my CombineGVCF done in step 3 can take anywhere from 4-8 hours. I was wondering if I could change this step to use CombineVariants and have the result be the same (unlost data). The main reason for this would be because GATK currently allow CombineVariants to use the -nt option.
Thanks for you time and work.
Referring to broadinstitute.org/gatk/guide/article?id=3060, is removing duplicates necessary to be done twice, once per-lane and then per-sample?
Is it not enough to just mark the duplicates in the final BAM file with all the lanes merged, which should remove both optical and PCR duplicates (I am using Picard MarkDuplicates.jar)? So specifically, in the link above what is wrong with generating -
Then, merging them to get
and finally, include "de-dupping" only for the merged BAM file.