NLOD and TLOD in MuTect2
I'm trying to understand better the rationale of the algorithm behind MuTect2 (GATK3.6) . My question are based on your slides from a workshop (16th of September 2016).
I just still cannot understand where the TLOD and NLOD are calculated and if they are linked to the PairHMM likelihoods. As far as I understood, the procedure is :
1 ) Finding active regions
Is there any likelihoods used for that ? Are TLOD and NLOD used here ? Or is it just some hard-filter to exclude certain regions?
2 ) Assembles plausible haplotypes
for now I prefere not to enter into the details of this...
3 ) Calculate PairHMM likelihoods
So for now what I understood : PairHMM calculates likelihoods ... and those likelihoods allow us to calculates the log-odd ratio TLOD and NLOD at the very end ..?
4 ) Choose the most probable genotype
As far as I understood the likelihoods calculated in 3) are used here to first choose the most probable genotype, if it is not the homozygous reference it is called variant...
What about NLOD and TLOD ? Is it just a supplementary hard filter after 4) ? Does it use PairHMM likelihoods? Are definitions of NLOD / TLOD the same as in MuTect1 ? (With another way of calculating the likelihood as far as I understood) , meaning :
NLOD : ratio of Somatic Mutation Likelihood model over Germline Mutation Likelihood model ?
TLOD : ratio of Real Variant Model likelihood over Noise Model likelihood ?
Thank you very much !