Periodic Reporting for period 1 - MacDrive (Which Factors Drive Macroevolutionary Rates of Speciation and Extinction)
Période du rapport: 2023-01-01 au 2025-06-30
The MacDrive projects aims to discover what drives macroevolutionary rates of species diversification. Specifically, we explore the impact of species-specific (e.g. habitat, diet and body size) versus shared factors (e.g. environmental or abiotic) on speciation and extinction rates. To achieve this goal, we work on time-calibrated phylogenies that include extant and fossil species. Since well-sampled species-level phylogenies including fossils do not exist yet or are constructed with ad-hoc fossil placement methods where branch lengths are not inferred in units of time, one of the major objectives of MacDrive is to infer such well-sampled species-level phylogenies of carnivora, cetartiodactyla, squaliformes and crocodyliformes. This inclusion of fossil species in phylogenies fundamentally relies on morphological data, as molecular data is extremely rare for fossils, and until today not available for fossils older than ~20 million years. Thus, our current work focuses both on constructing comprehensive morphological data matrices by scoring more characters for extant taxa and including more fossil species as well as developing more robust inference approaches and models for phylogenetic inference from morphological data. Overall, this project combines theoretical, computational and empirical work. Specifically, to address the overall go to discover what drives macroevolutionary rates of species diversification we (i) assemble new empirical datasets with a special focus on morphological data of extinct and extant species, (ii) develop new statistical models for inferring time-calibrated phylogenies with extinct and extant species, and (iii) build new stochastic models and inference approaches to infer species diversification rates using these phylogenies.
The first phase of the MacDrive project focused on three work directions: (i) assembling of new empirical datasets with a special focus on morphological data of extinct and extant species, (ii) develop new statistical models for inferring time-calibrated phylogenies with extinct and extant species, and (iii) build new stochastic models and inference approaches to infer species diversification rates using these phylogenies. We have currently successfully assembled molecular datasets for crocodilians and squaliforms. We used the molecular dataset of crocodilians to establish a robust phylogeny with divergence times. Ongoing efforts focus for these two datasets on the inclusions of fossil taxa based on morphological data. We have furthermore compiled a best practices and guidelines for inferring time-calibrated phylogenies with extinct and extant species. These best practices are the foundation of our ongoing work to estimate phylogenies with fossils as tips. Additionally, our best practices and guidelines have identified several shortcomings in specifically using morphological data. We have addressed some of these shortcomings by developing novel models for discrete morphological character evolution, and are currently continuing this effort with the aim to provide robust phylogenetic inference methods. Finally, we have developed novel inference procedures to estimate lineage specific diversification. This method will provide the foundation to estimate diversification rate which we then test for correlation with species-specific and/or environmental factors.
Our novel approach to infer lineage-specific diversification rates in a full-likelihood framework provides a major breakthrough as it enables application to large mega-phylogenies (>100k taxa) as well as computationally feasible model exploration. The conundrum with diversification rate estimation has been, that very large phylogenies are required to infer general patterns of diversification rate variation, but these large phylogenies cannot be handled by existing likelihood-based methods. Thus, with more available data (i.e. larger phylogenies) researchers had to resolve to less powerful approaches. Our fundamentally different approach that circumvents time-consuming Markov-chain Monte Carlo approaches provides the starting point for new types of diversification rate estimation methods. We see this work as a starting point for a new era of diversification rate inference methods, thus a major breakthrough for the research community.