Scalable Model-Based Reconstruction of Network Evolution

University Of Wisconsin-Madison, Madison WI

Investigators

Abstract

The availability of full genomes across individuals from many populations and many species offers rich information about past evolutionary history. By comparing the genes of different individuals, one can detect which individuals are most closely related and reconstruct the history of population splits and speciation, as visualized in a phylogenetic tree. Challenges arise however because of genealogical differences between individuals within each species, current or ancestral. This project focuses on the detection of species convergences: when species hybridize, or when individuals from one species migrate to another, or when strains recombine. The history of a group of species is then best described by a network, where a backbone tree represents speciation and extra branches describe gene flow from one population into another. Current methods to estimate phylogenetic networks cannot analyze data sets with more than a few dozen species. Based on novel theoretical foundations, the PIs will develop statistical methods and software that will scale to hundreds of species and thousands of genetic loci. These new methods will also be particularly valuable to advance knowledge in bacterial and virus evolution, where recombination is prevalent. The project will support graduate and undergraduate students, who will gain training beyond traditional disciplinary boundaries with involvement in the larger community of campus researchers interested in networks in data science. Through the mathematical analysis of coalescent processes on phylogenetic networks, the PIs will determine the maximal substructures of these networks that can be theoretically identified from various data types, such as from gene trees, or genetic distances between pairs of individuals, using one or more individuals per populations. Theory will also be developed to determine the amount of data necessary to reconstruct the phylogenetic network with accuracy. These theoretical findings will guide the development of new statistical methods and software to estimate phylogenetic networks from data, with a focus on the use of genetic distances to devise fast algorithms that can handle hundreds of species. These fast reconstruction methods will allow the deployment of a cross-validation method to learn from data the appropriate complexity of the network, that is, the appropriate number of gene flow events. The proposed research will advance knowledge of the evolutionary history in many groups where gene flow and recombination is suspected, such as the early radiation of mammals or land plants, and the evolutionary history of the herpes virus family. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.