Table 2 lists the hidden states used in our HMM. Note that we divide the normal (diploid) state into homozygote and heterozygote sub-states to take into account regions of homozygosity since the frequency of homozygotes in heterozygous regions (2/3) differs from that in homozygous regions (1/2). We use an exponential function to define an a priori probability that some genetic event (hidden state change) occurs between adjacent SNP loci a distance d apart, 1 where L is a characteristic length which could either be inferred directly from the data, or adjusted to calibrate the model to a given false positive rate in an objective fashion (see below). The transition matrix of hidden states between adjacent SNPs i, j is given by: 2 where h is the rate of heterozygosity which we set as 1/3 (chosen based on the mean AB frequencies given in the BeadStudio Manual), and Ns is the number of hidden states. Table 2.Hidden states, associated copy numbers and biological interpretationHidden state, zCopy number, c(z)Number of genotypes, K(z)Description100Full deletion211Single copy deletion323Normal (heterozygote)422Normal (homozygote)534Single copy duplication645Double copy duplicationWe associate
