Collaborative Research: Fast and sound coalescent-based inference of semi-directed phylogenetic networks from genome-scale data

A question that has fascinated humans for centuries is trying to understand evolutionary relationships between species. From such inquiries, scientists have determined that chimpanzees and bonobos are the apes most closely related to humans, and that butterflies regularly hybridize so that wing patterns like stripes and dots can be passed along from parents of different species to children who make up a novel species. The art of determining such kinship relationships using a tree or network dates back to Darwin and has current applications not only in evolutionary biology but in biomedical research and other fields. As an example, trees are used to visualize the relationship between viral strains like that causing COVID, as an aid in understanding disease spread and virulence. This research focuses on developing new mathematical and statistical methods for determining evolutionary relationships using current genome-scale data like full genome sequences or thousands of gene trees constructed from stretches of DNA that code for proteins. Emphasis is placed on soundly determining the history of hybridization, where genomic material from several parent species is combined in new ways and passed along to novel offspring. This project will develop new scientific theory for detecting hybridization, implement practical algorithms in easy-to-use software, and introduce statistical measures for assessing how well these methods work. To help in preparing the next generation of scientists, this project will run a summer school and develop teaching materials for others to use in learning state-of-the-art tree and network reconstruction methods. With genome-scale data readily available, there is a pressing need for fast and statistically sound methods for inferring networks depicting evolutionary relationships using these data as input. This project develops provably consistent, fast, distance-based semi-directed network and tree of blobs topology estimators under the Network Multispecies Coalescent Model (NMSC), using both sequence and gene tree data as input, and statistical support measures for features of these estimators. This is accomplished by (1) proving parameter identifiability theorems under the NMSC from summary distributions for higher-level and more complex semi-directed networks; (2) translating identifiability results into fast, scalable inference algorithms; (3) introducing consensus methods, support measures for graphical features of such networks and trees of blobs, and dissimilarity measures for pairs of semi-directed networks; (4) implementing algorithms in freely-available, user-friendly software; and (5) validating all new methodology on simulated and empirical data sets. Educational materials for training the next generation of phylogenetic researchers will be developed. This work combines biological modeling, statistical theory, combinatorial methods, and graph-based algorithms to improve the inference of phylogenetic networks so that researchers gain better understanding of reticulate processes, such as those occurring in evolutionary biology, stemmatology, or biomedical research. 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.