Fig. 2

Polyploidy, genome evolution and the POInT computation. A A polyploidy event followed by a speciation and duplicate gene losses produces regions of double-conserved synteny (DCS) in the two resulting genomes, raising the issue of phasing those regions with respect to each other. A region of five genes (brown) in the non-polyploid ancestor is duplicated (pink) and experiences three duplicate gene losses prior to a speciation event (red “X”s). After the speciation event, the two resulting species also experience independent losses, yielding the blocks of DCS in each genome. B There are 2ⁿ = 4 possible orthology relationships for the DCS blocks in these n = 2 genomes. These relationships are shown for the example (known) loss patterns from A. C For each orthology relationship in B, the likelihood of the observed presence/absence data at that pillar can be computed on the species phylogeny with a Markov model of duplicate losses. Those likelihoods can then be conditioned on the other pillars in the dataset. In this model, undifferentiated duplicate genes (U) can either be fixed (F) or lost from subgenome 2 (copy 1 or S₁ survives) or lost from subgenome 1 (S₂ survives). As the ε parameter (0 ≤ ε ≤ 1) approaches 0, subgenome 1 is increasingly favored over subgenome 2. Model parameters and tree branch lengths are estimated from the pillar data by maximum likelihood [9]. From this model, the relative likelihood of each of the orthology relationships in B, conditional on the full dataset, can be computed: these values are the confidence estimates at the top of the pillars in Fig. 1