Research


A network view of evolution

I subscribe to the view that a network inspired view of evolution is very useful. Under this view, genotypes, connected by mutations, form a network on top of which evolution proceeds. A population evolves by spreading across this network, according to mutation rates, population sizes and phenotypes and fitness associated with each genotype. Robustness and evolvability are properties of the genotype-phenotype map this network is embedded with. For example, mutational robustness of a genotype (the invariance of the phenotype in the face of mutation) is reflected by the number of neighbors that share the same phenotype (neutral mutations). Many insights from traditional population genetics can be recovered in this view in a very natural way, and many other insights can be gained by adopting this perspective.
Traditional evolutionary theory tells us how fitness changes but not how the individual adapts. One reason is that the concept of phenotype is well hidden in traditional population genetics, and mostly reduced to fitness. My aim is to bring phenotypes and their mechanistic basis back into evolutionary theory.
In fancy terms, I am generally interested in the evolution of variational properties of genotypes and phenotypes. More specifically, I am interested in questions about robustness and evolvability and how these properties evolve, i.e., the mechanistic basis of the evolutionary process. I am interested not just in the the evolution of specific phenotype producing mechanisms, such as gene networks, and how these mechanisms constrain or enable evolution, but also on general properties of the evolutionary process itself. The general questions I am trying to answer are:
  • How do specific mechanisms constrain or enable the production of phenotypic variability?
  • Are there generic properties of these genotype-phenotype maps? How much do they tell us about evolution?
  • Is the evolutionary process predictable? If so, can we predict it?
  • What are the limits to the complexity that adaptation can produce?
I am also easily sidetracked into working on anything I find genuinely interesting, like sex ratio evolution in Tetrahymena

Below you can find some examples of things I've been working on. You can also learn a bit more about I did for my PhD here.


Selection limits on correlated fitness landscapes

The efficacy of natural selection to produce complex adaptations has been a long-standing question. Are there limits to the complexity that natural selection can produce? Was there enough time for natural selection to produce the complexity observed in the natural world? These questions have been at the heart of populations genetics since its beginnings.
One particular difficulty has been that the speed of adaptation is believed to strongly depend on the details of the genetic architecture of the organism, which is largely unknown. Gene interactions and number of sites contributing to the trait are expected to have a strong effect on the time required to evolve a particular adaptation. Both of these factors are virtually impossible to measure in nature.
We make use of tools from theoretical computer science to directly address these difficulties. We present simple expressions for the time to reach a particular genotype as a function of the length of the target sequence and, instead of focusing on a particular landscape, we focus on large classes of fitness landscapes, thereby including many patterns of gene interactions. Particularly, we provide such scalings for all unimodal functions, addressing all patterns of gene interactions that do not create local peaks. Our results provide conditions under which natural selection is efficient and their dependence on the underlying genetic architecture.

The role of gene interactions on the response to selection

The role of gene interactions in the response to selection has long been a controversial subject; while some have dismissed them as an important influence on adaptation, others have argued that their long-term effects are of high significance. At the hear of this debate is the distinction between functional and statistical epistasis. The former relates to gene interactions while the former refers to a statistical description of the effect of these interactions on the phenotypic variance of the population. We derive simple and general predictions for the effect of gene interactions on the long-term response to selection in two extreme regimes. We show that when the dynamics of allele frequencies are dominated by genetic drift, the long-term response is surprisingly simple, depending only on the initial components of the trait variance, regardless of the detailed genetic architecture. In the opposite regime, when selection dominates the dynamics of allele frequencies, the long-term response depends only on the genotype-phenotype.

Evolution of gene regulatory networks

One of the insights brought by the genome project is that the complexity of organisms is not correlated with the number of genes. Instead, it is the combinatorial nature of gene regulation that seems to be important. Because of this, the evolution of interaction networks among genes became an important subject in biology.
“Redundancy and the evolution of cis-regulatory element multiplicity” introduces a model for the evolution of networks of transcription factors that accounts for realistic sequence based evolution of cis-regulatory elements. We illustrate the use of this model by investigating the role of many evolutionary and structural factors on the evolution of binding site redundancy. Among other things, we identify sexual reproduction as an especially important factor in the revolution of redundant forms of transcriptional regulation.
This paper integrates neutral network concepts with mechanistic modelling of biological processes. It illustrates the role of neutral networks in evolution and how redundant forms are essential for the evolution of changes in transciptional regulation. Moreover, our treatment of recombination in this model suggests a geometric method of integrating sexual reproduction with discrete mutational landscapes. 

Sex ratio evolution under probabilistic sex determination

Tetrahymena thermophila displays an unusual sex determination system. In typical genetic sex determination systems, the sex of the offspring is determined by their genotype. In T. thermophila, in contrast, a single locus determines not the sex but the probability of acquiring one of seven self-incompatible sexes. Motivated by this, we used a population genetics model to investigate the consequences for the evolution of sex ratio of this peculiar sex determination system. We found that, under limited genetic variability, this sex determination mechanism introduces genetic constraints that can lead to skewed sex ratios. Furthermore we found that probabilitic alleles, such as the ones found in T. thermophila, can outcompete deterministic alleles, raising the possibility that this sex determination system is more common than currently thought, at least among ciliates.
“Sex ratio evolution under probabilitic sex determination” investigates evolution under genotype-phenotype maps in which one genotype leads to many different phenotypes. 

The predictability of evolutionary trajectories

How does the evolutionary process proceeds? How are the population dynamics of an adaptive walk influenced by evolutionary factors like mutation rates, fitness and population size? Can we predict the order in which mutations are incorporated in a population? "what would happen if we replayed the tape?"
Here we address these questions by simulation adaptive walks in different mutational landscapes. We use these simulated walks to understand how different evolutionary parameters affect evolutionary trajectories. In order to test our power to predict evolution, we use population genetics to generate predictions about the population dynamics and contrast them to the simulated walks. In essence, we ask whether evolution is repeatable and if so, can we predict it?
We found that the evolutionary process is indeed repeatable with some trajectories much more likely than others. Furthermore, we find that clonal interference improved the repeatability of an adaptive walk. Counter-intuitively, however, we find that beyond a critical population size, larger populations become less predictable and that this can lead to dramatic shifts in the trajectories followed by the population.
Subpages (1): Phd Thesis