Presentation Detail

SSB Symposium: Ernst Mayr Symposium

Hoehna, Sebastian [1].

A new birth-death model recovers the K-Pg mass-extinction event in mammals.

Reconstructed phylogenetic trees from molecular data are frequently used to infer patterns of diversification, such as rapid radiations and mass-extinction events. However, until now the impact on the diversification of mammals of the Cretaceous-Paleogene (K-Pg) mass-extinction event could not be detected from reconstructed trees despite the fossil evidence. The reasons are model violations of the birth-death process, which is used for the inference. I will present several extensions to the birth-death process that enable the inference of the K-Pg mass-extinction event.First, most inferences have either assumed constant diversification (speciation and extinction) rates or used simplistic rate functions such as an exponential decreasing rate function. These assumptions are very unrealistic for large clades, as the mammals. Therefore, I implemented a Bayesian diversification rate skyline model with additional mass-extinction events to allow for complete variability of the diversification rates. Second, the birth-death process assumes either complete taxon sampling or uniform taxon sampling, i.e. each taxon has the same probability to be sampled. These assumption are clearly violated on the family level mammalian phylogeny constructed by Meredith et al. (2011). In a previous study we showed that such model violation can severely bias the diversification rate estimates. To overcome this problem I derived and implemented an extension to the current sampling strategies, namely age-dependent sampling, that incorporates information on how species were selected.The results based on the extended birth-death process clearly recovers the K-Pg mass-extinction event and non-constant diversification rates. Additionally, I tested the power to infer rate variation and mass-extinction events by simulations, which required new algorithms to simulate reconstructed phylogenies. The simulations verified the ability of the model and inference procedure to recover mass-extinction events. Goodness-of-fit of the models was tested using posterior predictive analysis. All methods and likelihoods are implemented in the R package TESS, which is freely available from CRAN.

Related Links:
R package

1 - Stockholm University, Department of Mathematics, Stockholm, Sweden

Keywords:
birth-death process
diversification
Bayesian inference
Mammals
mass-extinction.

Presentation Type: Symposium Presentation
Session: 212
Location: Ballroom 2/Cliff Lodge
Date: Sunday, June 23rd, 2013
Time: 11:15 AM
Number: 212004
Abstract ID:1206
Candidate for Awards:Ernst Mayr Award