Testing hybridization hypothesis within a multilocus coalescent framework: an example with Liolaemus lizards (Squamata:Liolaemini, boulengeri and rothi complexes)

OLAVE, M.; AVILA, L. J.; SITES, J., JR.; MORANDO, M.

Snowbird (UT)

Congreso; Evolution Meeting 2013; 2013

The field of phylogeography is experiencing rapid growth in many directions, and although scientists know the importance of making studies implementing multi-locus datasets (Maddison 1997), statistical and computational issues remain challenging (Than and Nakhleh 2009). One of the most recent approaches is Approximate Bayesian Computation (ABC, Beamount et al. 2002). ABC approaches bypass exact likelihood calculations by using summary statistics and simulations. Its? roots in the rejection algorithm consist in simulating large datasets under a hypothesized evolutionary scenario, and then contrasting with the observed data to found or reject support of that hypothesis (Csillery et al. 2010). This provides a large flexibility, and allows users to develop models that are not included by regular software used in phylogeography. This makes ABC a very good approach to test the hypothesis of gene flow between species during their speciation. Actually, many programs have been developed to simulate, as well as to carry out the rejection and regression steps (e.g. SIMCOAL [Laval and Ecoffier 2004], ms [Hudson 1983], MaCS [Chen et al 2009], msBayes [Hickerson et al 2007], DIYABC [Cornuet et al 2008], ONeSAMP [Tallmon et al 2008], ABC4F [Foll et al 2008], PopABC [Lopes et al. 2009], 2BAD [Bray et al. 2009]). Also, ABCtoolbox (Wegmann et al. 2009) has provided the opportunity of linking some of that software, helping users to work with an ABC approach. Even thought the flexibility of ABC allows modeling several evolutionary processes relevant for the particular case of the study object, it also presents some limitations, especially because computational limitations and time required for simulating DNA sequences and estimating summary statistics. However, an optional framework implies simulating gene trees given a model, which works significantly faster. Here we developed a function, written in R (R Core Team; http://www.R-project.org), loads a given model, calls ms software and returns a p-value of observing the real gene tree under that particular model. We look forward to use it to test the hypothesis of gene flow using boulengeri and rothi complexes of Liolaemus lizard genus.