Probabilistic Inference of Evolutionary Gene Branching Events

A method for finding probable phylogenetic gene branchings is presented. The events that caused the branchings are estimated by using probabilistic orthology analysis. The orthology analysis is based on a model for gene evolution in a species phylogeny. Markov chain Monte Carlo sampling is used to sample gene phylogenies from an a posteriori distribution after having observed gene sequence data. The samples are then used to find branchings with high probability. Any branching is seen as the start of two independent gene lineages. Possible events causing gene branchings are grouped into two complementary classes: speciation events and gene-duplication events. To estimate which event caused a branching, an approximation of the orthology probability among extant genes from the independent lineages is used. Branchings with high probability are then used to construct a consensus tree for visual analysis. Each branching is annotated with its branch-event estimation. Finally an experiment is performed to determine the accuracy in the branching-event probabilities. The results indicate that the method has some problems correctly classifying speciation events, but most duplication events are correctly classified. However, more experiments are needed to acertain these conclusions.