Bioinformatics Leipzig

[ PDF ]

Alg. Mol. Biol.

BACKGROUND: Coevolution between species is a common phenomenon in biology: species interact across groups such that the evolution of a species from one group can be triggered by a species from another group. Most prominent examples are systems of host species and their associated parasites. Typically in this field, phylogenetic trees for both groups of species can be constructed from sequence data or/and morphological data. In addition, the host parasite interactions between the extant taxa are known empirically. The problem is then to reconcile the common history of both groups of species and to predict the associations between ancestral hosts and their parasites. Some algorithmic methods have been developed in recent years to solve this reconciliation problem. Only few host parasite systems, however, have been analyzed in sufficient detail to serve as benchmarks for the evaluation of the reconstruction methods. RESULTS: We propose to tackle the lack of benchmarks by generating reliable test data sets with a dedicated approach for generating cophylogenies. Our method builds on biologically motivated branching models to generate cophylogenies under the assumption of the widely used coevolutionary model. It pictures coevolution as a stochastic process with cospeciation, duplication, lineage sorting and (host) switching as discrete events. The probability of an independent parasite speciation as well as the ratio between cospeciations and sortings and between duplications and host switches are user defined parameters. We evaluate choices of reasonable parameter settings under the aspect of producing realistic coevolutionary scenarios, giving rise to a large set of test scenarios. Based on these scenarios, we provide a detailed analysis and comparison of the common reconciliation tools TreeMap 3b, Jane 2.0, and CoRe-PA with a focus on the significance of the computed reconstructions. All three tools are based on the maximum parsimony principle but using different heuristics and cost models. To the best of our knowledge, this is an initial contribution to extensively compare methods for cophylogeny reconciliation. The stochastic model is the first evaluated method to obtain cophylogeny evaluation data sets in a controlled manner. CONCLUSIONS: Out of the three tools applied to the test data sets, CoRe-PA yields the most precise predictions of the associations between hosts and parasites. However, Core-Pa does not optimally estimate the number of cospeciation and switching events. Furthermore, CoRe-PA is the computationally most expensive method. Jane 2.0, being the fastest of the three tools, is best at estimating the correct number of cospeciations. A drawback of Jane 2.0 is its dependence on a single user specified cost model. TreeMap 3b is the only tool with the option to find the optimal reconstruction for a specified cost model. In terms of accuracy of the computed reconstructions, TreeMap falls short of the other tools.