Abstract

From many data types, we hope identify the phylogenetic semidirected network that generated the data, but not the rooted network. In the semidirected network, some but not all edges are directed. I will define a general class of semidirected phylogenetic networks, with a stable set of leaves, tree nodes and hybrid nodes. This class includes both rooted phylogenetic trees and networks, and unrooted phylogenetic trees. I will show various fundamental properties for these networks, including how to generalize the ""tree-child"" property from rooted to semidirected networks. I then will define an edge-based representation of semidirected phylogenetic networks, which generalizes the node-based μ-representation of a rooted network by Cardona et al. (2009), and the split-based representation of an unrooted tree. It leads to a dissimilarity between semidirected networks, which can be efficiently computed in near-quadratic time, and extends the widely-used Robinson-Foulds distance on both rooted trees and unrooted trees. This dissimilarity is in fact a distance on the space of tree-child semidirected phylogenetic networks.

• Video Available Soon

Abstract

Gene flow is a highly dynamic suite of processes that includes events like lateral gene transfer, hybridization, and introgression. Even between clades across the Tree of Life, these processes can have different controlling factors and mechanisms that ultimately affect the macroevolutionary patterns we observe. In this talk I will discuss some of the ways that various gene flow processes can be modeled on a macroevolutionary scale. I simulated phylogenetic networks from the birth-death-hybridization process to demonstrate that phylogenetic networks can have drastically different diversification dynamics and topological profiles as we change our underlying assumptions around reticulate patterns. This work emphasizes the need to think critically when choosing an appropriate and fitting model for a given empirical system. This marks an exciting opportunity for biologists to use their system-specific knowledge to model gene flow appropriately and test macroevolutionary hypotheses regarding the role of gene flow to affect diversification and adaptation.

Abstract

Modeling of trait evolution usually assumes a known phylogeny on which traits evolve such as from Brownian Motion or an OU process. It is conceptually not clear if the phylogeny should be a species tree or a gene tree. Here, we develop a model where a trait evolves on two different gene trees. We introduce a parameter p for the contribution of each gene so that a researcher can estimate the proportion of contribution from each. The example is applied to cultivated rice with two genes that strongly influence grain width but have very different topologies.

Abstract

Evolution is a process that involves an organism’s DNA changing over time. Often, this process is driven by external stressors (such as environmental changes, diseases, pests) and/or selective breeding. Recognizing how genomes evolve is of universal biological interest about species origin, survival and adaptation. The fast-increasing number and diversity of sequenced genomes open the door to analyzing the dynamics of evolution at whole-genome resolution. By using a successive ancestral contig construction technique based on statistically conserved syntenic orders of genes, we developed methods to reconstruct chromosome-level ancestral genomes of extant species or subgenomes within a single genome as a result of whole genome doubling/tripling. These methods were applied and verified in many lineages across different studies that improved the scientific knowledge of genome evolution in various biological systems. In the talk, I will first introduce RACCROCHE and SyntenyLink, pipelines for ancestral genome or sub-genome reconstruction, and our research on genome evolution based on genes, their syntenic orders, chromosomal gene co-occurrences, subgenome dominance in gene expressions, novel genomic variations (structural variants and transposable elements) and their relationship with plant phenomics. A particularly important application of this work is to crop species that exhibit rapid rates of evolution. By understanding genome evolution and its impact on plant phenotypes, breeders will be able to make informed decisions on how to breed crops with desired traits.

Abstract

Is it possible to reconstruct the correct phylogenetic network if one is given sufficient data? This question has statistical, algebraic, combinatorial and algorithmic sides and has gained quite some interest recently. In this talk, I will discuss recent progress and major open questions related to the identifiability of semi-directed phylogenetic networks under Markov models, the usefulness of quarnets and quadratic phylogenetic invariants, algorithms for constructing the tree of blobs and for constructing level-1 networks, and the elusiveness of level-2.

Abstract

The development of a multicellular organism from a single founder cell is the result of a complex process involving growth, death, and differentiation of cells. Recent advances in genome editing and single cell sequencing technologies allow researchers to induce somatic mutations in the genome during development and to simultaneously measure induced mutations and cell state from thousands of cells. In this talk, I will describe two algorithms to infer the developmental process from such dynamic lineage tracing data. First, I will describe Startle (Star tree lineage exploration), an algorithm that reconstructs cell lineage trees from mutations induced during genome editing. Startle models the unique characteristics of the genome editing process using the star homoplasy model to overcome errors and missing data in the measured mutations. Next, I will describe our algorithm Carta to infer a cell differentiation map – a directed graph describing transitions between cell types – from single cell lineage tracing data. Carta models unmeasured progenitor cell types using the concept of a potency set and infers a cell differentiation map that balances the trade-off between the complexity of the cell differentiation map and the number of unobserved cell type transitions on the lineage tree.

Abstract

The utility of whole-genome sequencing for pathogens, and how informative sequence data are for outbreaks, remains unknown, and is important as public health systems around the world decide how much sequencing to undertake for infectious disease surveillance. In this talk, I introduce an MCMC method for simultaneous reconstruction of phylogenetic trees and transmission trees using sequence data for a person-to-person outbreak of an infectious disease. The method is called BREATH: Bayesian Reconstruction and Evolutionary Analysis of Transmission Histories. BREATH's transmission process accounts for a flexible natural history of infection (including a latent period if desired) and a separate process for sampling. It allows for unsampled individuals and for individuals to have diverse within-host infections. BREATH also accounts for the fact that an outbreak may still be ongoing at the time of analysis, using a recurrent events approach to account for right truncation. We perform a simulation study to verify our implementation, and apply BREATH to a previously-described 13-year outbreak of tuberculosis. We find that using a transmission process to inform the phylogenetic reconstruction results in better resolution of the phylogeny (in topology, branch length and tree height) and a more precise estimate of the time of origin of the outbreak. But even with whole-genome sequence data, considerable uncertainty remains about transmission events. An open source implementation of BREATH is available from https://github.com/rbouckaert/transmission as the BREATH package to BEAST 2.

Abstract

Hybridization plays a crucial role during the evolutionary process of some species. In such cases, phylogenetic trees are insufficient to describe species-level relationships. Phylogenetic networks (also known as admixture graphs in population genetics) are the objects used to represent the evolutionary histories that allow such interactions. The inference of phylogenetic networks from biological data presents many challenges, and methods have been developed under different frameworks and model assumptions to overcome some of these challenges. In this talk, we present a comprehensive overview of existing methods for estimating explicit networks (networks depicting evolutionary history) from biological data. We discuss their distinct model assumptions, framework, hypotheses, and drawbacks.

• Video Available Soon

Abstract

Synesth is one of the most comprehensive and flexible tool for tree reconciliation allowing for events on syntenies (i.e. homologous genomic regions containing multiple genes evolving together) including duplications, transfers, fissions, and transient events going through unsampled species. This model allows building histories that explicate the inconsistencies between a synteny tree (i.e. with leaves representing syntenies rather than single genes) and its associated species tree. An inductive characterization of the solution space using different algebras for each granularity (Pareto-optimal histories, Pareto-optimal event count vectors or minimum event cost histories) leads to efficient dynamic programming algorithms. We show how Synesth can be applied to infer evolutionary scenarios for CRISPR-Cas systems. We address the challenge of building a synteny tree and present FullSynesth, a generalization of Synesth rather starting from a set of gene trees. Finally, we show how this field opens the way to many unexplored algorithmic problems.

• Video Available Soon

Abstract

Birds are an extensively studied clade with more than 10,000 named species. Many relationships within birds have proven to be difficult to recover in phylogenetic analyses. This difficulty reflects several factors, including discordance among gene trees, shifts in evolutionary rates, changes in the model of sequence evolution across the avian tree, and the complexity of the recombination landscape for bird genomes. There are multiple efforts to generate sequence data for all bird taxa; this talk will focus on an effort to use ultraconserved element (UCE) sequence capture to generate data for all currently recognized bird species. Analyses of UCE data for large numbers of taxa presents several challenges. More specifically, I will describe the performance of phylogenomic methods with a UCE dataset with hundreds and thousands of taxa. The methods used in these comparisons include the commonly used methods (maximum likelihood analysis of concatenated data, gene tree summary methods, and SVDquartets) as well as a “new” method called METAL (Metric algorithm for Estimation of Trees based on Aggregation of Loci). Empirical studies have obvious limitations since they represent a single realization of an evolutionary history, but they also provide information that is difficult to obtain in more controlled situations like simulations.

Abstract

Inference of phylogenetic networks from biological sequence data is a challenging problem, with many theoretical and practical questions still unresolved. In this talk, we introduce a novel divide-and-conquer approach for inferring level-1 networks under the network multispecies coalescent model. We first infer the “tree of blobs” of the network, in which all reticulations are collapsed into nodes. We then focus on resolving the multifurcations within this tree, which correspond to the network’s cycles, by determining an optimal cycle structure for each. Once the multifurcations are individually resolved, we combine them to obtain the full network, provided the cycles collectively allow for a rooting to exist. This method will be implemented in the R package MSCquartets, and we demonstrate its effectiveness through both simulated and empirical datasets. Additionally, we discuss some open problems and avenues for future research. This talk is based on joint work with Elizabeth Allman, Hector Baños, and John Rhodes.

• Video Available Soon

Abstract

Inference problems on gene-orders have suffered limited applicability due to complex models, and complications related to preprocessing the data. For instance, we lack practical ways to sample rearrangement scenarios, and traditional estimates of breakpoint re-use are highly dependent on the resolution of the syntenic blocks. We first simplify the task of counting scenarios by relating genome rearrangement scenarios to quadragulations and planar trees of a convex polygon. Next, we present an optimization problem for computing breakpoint re-use on a phylogeny, rather than between a single pair of genomes.

Abstract

One of the defining properties of a (rooted) phylogenetic network is that it has a unique root. For some biological processes such as introgression between lineages where it is important to preserve lineage information, this uniqueness of the root can make a phylogenetic network too restrictive a model to appropriately account for such processes. In the form of a forest-based network and an aboreal network, this has recently led to the introduction of a multiple rooted network to represent molecular evolution. In this talk, we will first introduce these types of networks and then present some novel results for them.

Abstract

Unrooted phylogenetic networks are often used to represent evolutionary data when incompatibilities are present. Although rooted phylogenetic networks are better suited for explicitly depicting evolutionary histories that involve reticulate events, they have been rarely used in practice, due to a lack of appropriate methods for their calculation. Here we introduce PhyloFusion, a new fast and easily-applicable method for calculating rooted phylogenetic networks on sets of rooted phylogenetic trees. The algorithm can handle tens of trees on hundreds of taxa, allowing unresolved nodes (which arise when edges with low support are contracted) and missing taxa. We demonstrate how the method can be used to explore the phylogenetic relationships between different Influenza A isolates, based on gene trees built on different genome segments.

• Live Stream

Abstract

Ancestral sequence reconstruction and phylogenetic reconstruction from present-day states are two important and related problems in computational biology. Multiple sequence alignment is a tertiary task that makes important connections between these two problems. Here, we show that perfect sequence alignment with high probability is possible with a dense enough underlying phylogeny influencing the evolution of genetic sequences. New probabilistic techniques in the evolution of sequences with insertions and deletions are developed, with a view toward more general state spaces. Based on work with Louis Fan, Christine Heitsch, and Sebastien Roch.

• Live Stream

Abstract

The advent of single-cell sequencing provides the ability to model the evolution of somatic cells within individuals. For example, inference of cell phylogenies has the potential to advance our understanding of the variation in the process of tumor progression. Phylogenetic methods have been applied in numerous ways to model the evolution of somatic cells from single-cell DNA data using single-nucleotide variants. Some methods implement a finite-sites Markov model by grouping genotypes in various ways, and such models are the focus of this presentation. Since both the maternal and paternal genomes are passed on to daughter cells during mitosis, it is appropriate to start with a phased model of maternal and paternal sequence evolution using a continuous-time Markov process. However, phased genotypes are not always observed, and only unphased genotypes are available. As a result, it is common to group pairs of heterozygous genotypes into single states, leading to an unphased model. Some methods further lump unphased genotypes, resulting in coarser processes. The process with a smaller state space corresponds to a lumped process of the original phased Markov chain and may or may not be Markovian, depending on restrictions imposed on the generator matrix. The purpose of this presentation is to explore the mathematical formulations of phased and unphased Markov processes. We hope to initiate a discussion about whether restrictions should be placed on the generator matrix when inferring tumor phylogenies in practice.

• Live Stream

• Live Stream

Abstract

Sequential Monte Carlo (SMC, particle filtering) has been used successfully to sample from Bayesian posterior distributions of phylogenetic trees. Our work seeks to apply SMC to the hierarchical Bayesian Multispecies Coalescent (MSC) model. In MSC, the primary distribution of interest is the marginal species tree posterior, which is challenging because the species tree is a hyperparameter that constrains the topology and node heights of the gene trees nested within and is thus only indirectly influenced by the observed sequence data. I will discuss our approach to applying SMC to the MSC model and illustrate with results from simulated as well as empirical datasets.

• Live Stream

Abstract

A central task in evolutionary biology is the construction of phylogenetic trees to represent the ancestral history of a collection of present-day taxa. The data typically used for this task are characters, which describe the attributes of the taxa under consideration. To illustrate, a particular position of aligned DNA sequences is an example of a character. The attributes (states) of this character are the subsets of taxa taking a certain nucleotide at this position. In the context of this task, many questions naturally arise. How do we decide if a collection of characters is compatible (realisable by a single tree)? Does compatibility depend on considering all of the characters in the collection simultaneously or can we do better than that? How many characters does it take to uniquely determine a phylogenetic tree? Does the answer depend on the topology of the tree? If so, how? If the characters have a bounded number of states, how does the number of characters it takes grow with respect to this bound and the number of taxa? In this talk, we investigate these questions, and discuss some old and not so old results.

Abstract

A model of genomic sequence evolution on a species tree should include not only a sequence substitution process, but also a coalescent process, since different sites may evolve on different gene trees due to incomplete lineage sorting. Chifman and Kubatko initiated the study of such models, leading to the development of the SVDquartets methods of species tree inference. A key observation was that symmetries in an ultrametric species tree led to symmetries in the joint distribution of bases at the taxa. In this talk, we explore the implications of such symmetry more fully, defining new models incorporating only the symmetries of this distribution, regardless of the mechanism that might have produced them. The models are thus supermodels of many standard ones with mechanistic parameterizations. We present phylogenetic invariants for the models, and discuss identifiability of species tree topologies using them.

Abstract

Placing a Markov model of evolution on a phylogenetic network gives rise to a statistical model that is particularly amenable to algebraic study by representing it as an algebraic variety. For group-based Markov models, these algebraic varieties are t-varieties with rich combinatorial structure. In this talk, we will present recent results regarding the dimension and defining polynomials of these algebraic varieties. We will then discuss how these properties can be used for establishing identifiability and network inference.