Fig. 8.

An example where MiniNJ computes distances that better reflect the true species tree than NJst. (a) The true rooted species tree. (b) A GFT resulting from a duplication at the root of the species tree. (c) The distance matrix $D_{\mathrm{NJst}}$ computed with NJst, incorrectly suggesting that all species are equidistant, except for C and D. This is the result of distance overestimation due to paralogous genes: for instance, species A and B are neighbors in the species tree, but the genes A2 and B1 are very distant from each other in the gene tree, because they start diverging at an early duplication event (paralogous genes). (d) Distance matrix $D_{\mathrm{MiniNJ}}$ computed with MiniNJ. The gene internode distances correctly reflect the species distances, because MiniNJ successfully pruned pairs of paralogous genes, such as A2 and B1, and only takes into account orthologous genes, such as A1 and B1.