Filtering methods for samtools or GATK

Question

[kot3]

Hello everyone!

I just started my PhD and my level of expertise is fairly low about bioinformatics! Is there any advice regarding different covarege levels?( 10x vs 50x or more ). I would need simple instructions for filtering using either samtools or GATK. We are looking for ROH (runs of homozygosity) so we need to avoid variants being called where they are not genuine. I tried to use MQ < 40 and DP>200 but not sure are they right or fixed.

Thanks for your answer!

Kardy

Answer 1

[LChart]

My gut instinct is that (after filtering at the variant level for artifacts to remove false-positives, mapping artifacts, etc) that you should not filter on the basis of depth, but instead use a Hidden Markov Model to assign states to regions of the genome. This will enable the use of genotype likelihoods in place of hard genotype calls. A framework for doing so was recently published: https://pubmed.ncbi.nlm.nih.gov/34923717/ https://github.com/gokalpcelik/ROHMMCLI