SNVMix1 is shown as a probabilistic graphical model in Figure 1C. The conditional probability distributions for the model are given in Figure 3 and the description of all random variables is listed in Table 1. The input is composed of allelic counts from aligned data and the output of inference is the predicted genotypes. Consider Gi=k, k∈{aa, ab, bb}, to be a Multinomial random variable representing the genotype at nucleotide position i, where aa is homozygous for the reference allele, ab is heterozygous and bb is homozygous for the non-reference allele. At each position, we have an observed number of aligned reads Ni. We let aij∈{0, 1} represent whether or not read j∈{1,…, Ni} matches the reference at position i. We let ai (no j index) be the total number of reads that match the reference at i. We assume the following likelihood model for the data: (1) where μk is the parameter of a Binomial distribution for genotype k. μk models the expectation that for a given genotype k, a randomly sampled allele will be the reference allele. Intuitively,
