A Dirichlet prior is used for the B allele frequency mixture weights, 6 where we set a strong prior on equal weights vw,z,k = 10 000, since we expect the relative frequencies of each genotype to be approximately equal although some departure is allowable if there is strong evidence from the data. This also prevents mixture component weights from collapsing to zero which would cause ambiguity, for example, in the normal state there should be three components, if one component were to disappear, there would be no difference in the emission distribution from a deleted state which has two.
