Extended Data Fig. 5: Posterior decoding from cobraa and cobraa-path, on simulated data.

We can decode the HMM of cobraa-path to infer the admixed regions of the genome (that is parts of the genome where the ancestral lineage pair went through AB or BB). The top panel of a shows the simulated lineage path across the genome, where the x-axis indicates the chromosomal position and the y-axis the ancestral lineage path. A structured model was simulated with μ/r = 1.25, 40% admixture, and constant population sizes. Using the simulated structured parameters, the middle panel shows the marginal posterior probability of each lineage path (from the forward/backward algorithm), and the bottom shows the most likely lineage path (from the Viterbi algorithm). (b) The full cobraa decoding of the simulation (hidden states are discretized coalescence times), where the y-axis indicates the coalescence time with 0 being the present. The green, dashed, horizontal lines indicate the simulated split and admixture times. (c) The full cobraa-path decoding of the simulation (hidden states are discretized coalescence times and ancestral lineage path). The y-axis indicates not only the coalescence time but also the ancestral lineage path, which is indicated by the shading in the right-most column. Red indicates AA, blue BB, and green AB. In the structured period, a red-blue pair indicates the same coalescence time, and more anciently than the split time a red-blue-green triple indicates the same coalescence time. The simulated states through the model are shown by the highlighted blue cells, and the posterior probabilities are indicated by the shading of the heatmap, with cream representing total confidence and black indicating no confidence. The y-axis in (a) indicates the ancestral lineage path, and in b and c the coalescence time, with more ancient states at the top and more recent ones at the bottom. The horizontal, green, dashed lines in b and c indicate the split and admixture time.