My main research interest lies in comparing theoretical models with empirical data, often using Approximate Bayesian Computation, a technique that relies on the ability to simulate data, and use the similarity between the simulated and the empirical data to obtain parameter estimates for the model of interest.
Hybridization dynamics on the genomic level
Hybrids of two Cottus species (C. rhenanus & C. parafretum) have only been relatively recently been formed, as the natural rivers in which these species occur (the Rhine and Scheldt river) were connected through canals about 200 years ago. Not only is this hybridization very recent, it turns out that the hybrids produced are actually fit, and are invading novel habitats previously uncovered by Cottus (such as the IJsselmeer and the Maas). When we look at the genomic level at these hybrids, we observe that the genome of the hybrids consists of a healthy mix of both parent species, although some alleles are more leaning towards one parent than the other. In this project we aim to identify to which extent selection acts upon certain genes, or whether variation in allelic frequencies are the result of drift.
Trait-based community assembly
Assembly of a community can be driven by many different factors, but here we focus mainly on trait-based effects, e.g. do the traits of species determine whether or not we observe a species in a plot? On the one hand we expect that species with traits that are very unfit for the local circumstances are filtered against (habitat filtering). If habitat filtering is an important process in trait-based community assembly, we expect the resulting trait distribution to be relatively narrow (as extreme values are filtered out). Conversely, if species have similar traits, they might perceive more negative interactions, for instance in the form of competition for food, or shared predation/parasitism pressures. If such limiting similarity is an important factor, we expect the trait distribution to become relatively evenly distribution and fairly wide. Lastly community assembly might be irrespective of traits, and we then expect the trait distribution to show no patterns at all. Starting out with a large species pool consisting of all species that could possibly migrate towards our focal community, we can start to select species that make it to the community based on either of these three processes. Using Approximate Bayesian Computation we can estimate the relative contribution of these three processes, given a species pool, the traits of these species, and the local community abundances.
How can we reconstruct the evolutionary history of species? Typically we do this by using the differences in genetic make-up of different species, and convert these differences into the amount of time spent separated (e.g. the time to the most recent common ancestor). Now that we have some kind of distance measure (in time) between species, we can use this distance matrix to construct a phylogenetic tree, and visualise the relationships between these species. Interestingly, the resulting phylogenetic tree can then subsequently be used to infer past speciation, and extinction, rates. In order to do so we can construct a simple model that describes how branches of a tree give rise to other branches, or die out (e.g. a birth-death model). For such models it is easy to formulate the full likelihood, and subsequently we can estimate birth and death rates from a given phylogeny. This does put an upper limit on the complexity of the birth-death model that is applied, as more complex models become analytically untractable. We might however want to implement birth rates that depend on geographical changes, on local diversity or on the age of the species. To circumvent this I apply Approximate Bayesian Computation Techniques to apply more complex models to given phylogenies.