Figure 2

Schematic diagram of cnvHiTSeq model. (a) State diagram of the haploid HMM used by cnvHiTSeq. Vertices represent haploid copy number states (0, red; 1, gray; and 2, green) and edges represent transitions from one state to another. The thickness of the edges is proportional to the transition probabilities of the model. More haploid states can be used to model higher copy numbers. (b) By pairing two haploid copy number states cnvHiTSeq derives the final diploid model (0, dark red; 1, red; 2, gray; 3, green; 4, dark green). (c) Each observed variable is modeled using a customized emission distribution. Here, as an example, we present the emission distributions for read depth and read pair distance conditional on the copy number state. Read depth is modeled using the negative binomial distribution, while read pair distance is modeled using the normal distribution. The color scheme corresponds to the diploid copy number states presented in (b). (d) The trained haploid HMM that corresponds to the homozygous deleted region on Figure 1a-c. The rows represent copy number states and the columns represent genomic location. Positions are given in megabases (Mb). Bubble size corresponds to the probability of assignment to each state at each position. The lines between bubbles indicate possible transitions between states. The annotated positions indicate the start and the stop of the CNV call. (e) The trained haploid HMM that corresponds to the heterozygous duplicated region on Figure 1d-f.