|
"Evolutionary biology has traditionally been a retrospective discipline, focusing on the history of life on Earth. However, the rapid evolution of many extant organisms, such as microbial pathogens in response to drugs or ecosystems in response to climate change, has motivated the need to make evolution a predictive, quantitative science. Can we predict the future evolution of these organisms?
Before making evolutionary predictions, we first must understand how predictable evolution even is. Since evolution inherently involves random processes such as mutation, its predictability is typically assessed by its repeatability: if we re-ran the evolution of a population many times, would we see the same outcome over and over again, or a different outcome every time? The answer to this question strongly depends on the biological scale at which we track the evolutionary outcomes. For example, we typically see a wide range of evolved nucleotide sequences in the genome and a narrow range of evolved traits of whole organisms. On the other hand, the evolution of nucleotide sequences contains much more information about the system than organismal traits do. This raises the question of whether there is an optimal scale for tracking evolution, which balances its repeatability (making it more predictable) and the information content of the data. For example, tracking evolution at an intermediate scale, such as protein abundances across the cell, may better balance repeatability and information content compared to tracking DNA sequences or organismal traits.
We have recently developed an information-theoretic framework to quantify the repeatability of evolution across biological scales. By calculating the information gain of an observed distribution of evolutionary outcomes relative to a prior expectation (i.e., how surprising the observations are given our expectation), we can determine the optimal scale for tracking evolution as the one that maximizes that information gain. We are testing this framework using three complementary approaches. First, we are using a model of protein evolution that describes how the amino acid sequence of a protein maps to organismal fitness via intermediate scales of protein biophysical traits (e.g., folding stability and binding affinity), cellular abundance, and metabolic flux. The model allows us to simulate large numbers of replicate evolution experiments and track the information gain at the different biological scales represented in the model. Second, we are developing abstract models of genotype-phenotype maps, inspired by neural networks, that generalize results of the protein biophysical model. And third, we are empirically testing the framework and the hypothesized optimal scale using genomic data from massively-parallel evolution experiments in Escherichia coli and Saccharyomyces cerevisiae and from a new experiment, carried out by our collaborators, on the evolution of antifungal resistance in Candida yeast.
The project is led by a postdoctoral fellow in my lab, Dr. Ernesto Berríos-Caro."
|
Sign in
to view more information about this project.
