Algorithms for Detecting Complete and Partial Horizontal Gene Transfers: Theory and Practice

We describe two methods for detecting horizontal gene transfers in the framework of the complete and partial gene transfer models. In case of a complete gene transfer model a new fast backward selection algorithm for predicting horizontal gene transfer events is presented. The latter algorithm can rely either on the metric or on the topological optimization to identify horizontal gene transfers between branches of a given species phylogeny. In case of the topological optimization, we use the well-known Robison and Foulds (RF) topological distance, whereas in case of the metric optimization, the least-squares (LS) criterion is considered. We also formulate and prove the NPhardness of the partial gene transfer problem. Second, an efficient algorithm for predicting partial transfers, using the Gauss and Seidel optimization, is discussed. We also show how to assess the reliability of a specific gene transfer or a whole gene transfer scenario. In the application section, we apply the new algorithm to detect possible gene transfers occurred during the evolution of