Should I use dbSNP to model known variants in BQSR in a random mutagenesis project?
I am currently processing some aligned reads to discover novel variants from an ENU mutagenesis project. All the samples are Bl6 mice, with a few exceptions.
A previous forum has discussed about using BQSR on mice sequences (gatkforums.broadinstitute.org/gatk/discussion/1243/what-are-the-standard-resources-for-non-human-genomes). Moreover, there's a research paper that fed in dbSNP database into their BQSR algorithm for their ENU project (ncbi.nlm.nih.gov/pmc/articles/PMC4623266/.)
From what I understand about BQSR, the recalibration method considers variants not found in well annotated variants as "errors" (is this error modelling?). I'm just wondering for such a project where novel mutations are to be expected, one might overestimate the errors if one were to feed in known variants into the learning algorithm (especially if the mutation load is high)?
Thanks in advance