A reconciliation between a gene tree and a species tree depicts an evolutionary scenario of the homologous genes in terms of gene duplications and gene losses. To infer such a reconciliation given a gene tree and a species tree, parsimony is generally used according to the number of gene duplications and/or losses. The combinatorial models of reconciliation are based on probabilistic or combinatorial criteria.;Based on the classical birth-death process, an algorithm that computes the likelihood of a reconciliation has recently been proposed. The second paper uses this algorithm together with the combinatorial tools described above to compute efficiently, either exactly or approximately, the posterior probability of the reconciliations located in the considered subspace. Based on realistic gene duplication and loss rates and on real/simulated datasets of fungal gene families, our results suggest that the probability mass of the whole space of reconciliations is mostly located around the most parsimonious ones. In the context of posterior probability approximation, our approach is a valuable alternative to a MCMC method and can competes against a sophisticated, efficient, and exact computation of the probability of a given reconciliation.;The Gene Tree Parsimony (GTP) problem is to infer a species tree that minimizes the number of duplications and/or losses over a set of gene family trees. Based on a new approach that explores the whole species tree space for the considered taxa and an efficient computation of the reconciliation cost, the third paper describes a Branch-and-Bound algorithm that solves exactly the GTP problem. When the considered number of taxa is too large, our algorithm can naturally take into account predefined relationships between sets of taxa. We test our algorithm on a dataset of eukaryotic gene families spanning 29 taxa.;Keywords: Gene family, gene duplication, gene loss, homology, comparative genomics, species tree, gene tree, coevolution, probability, reconciliation, parsimony, evolution and phylogenetic.;The first paper defines a simple and more general combinatorial model of reconciliation which clearly identifies duplication and loss events and does not only induce the most parsimonious reconciliation. An architecture of all possible reconciliations is developed together with efficient algorithms (that is counting, randomization, and exploration) to study combinatorial properties of the space of all reconciliations or only the most parsimonious ones.
展开▼
