Phylogenetics plays a key role in uncovering the evolutionary relationships among biological entities and in piecing together how the global biodiversity of life on Earth has developed over time. Traditionally, phylogenetic trees are used to represent ancestral relationships. However, processes such as horizontal gene transfer and hybridization result in relationship patterns that cannot be represented by a single tree. Indeed, phylogenetic networks are now widely accepted to better represent evolutionary histories. While several practical network inference methods have recently been developed, these methods are severely limited in their ability to scale up to large datasets and they require structural constraints that prohibit them from representing the true range of biologically realistic scenarios. In contrast, researchers have a precise understanding of the combinatorial structure of much broader classes of phylogenetic networks and have made progress in establishing combinatorial encoding results and associated algorithms that form the theoretical foundation to reconstruct such networks from smaller building blocks.

By synergizing combinatorial and statistical approaches, this Dagstuhl Seminar aims to establish new theoretical foundations that enable the development of robust and efficient software for inferring and analyzing phylogenetic networks. To this end, the seminar wants initiate new and strengthen current collaborations between two research communities that typically do not interact: combinatorial and statistical phylogenetics.

Topics to be discussed at Schloss Dagstuhl include:

1. Scalable algorithms for phylogenetic networks. Computationally efficient inference of phylogenetic networks for complex evolutionary histories remains out of reach for the large-scale datasets that are now commonly available. We will investigate how combinatorial encoding results can be combined with statistical inference methods to design scalable algorithms.
2. Identifiability of phylogenetic networks. To date, most identifiability results are restricted to simple classes of phylogenetic networks. In contrast, the combinatorial properties of more complex classes are well understood. We will explore for which classes of phylogenetic networks and evolutionary models identifiability results can be established and how these results can be translated into practical algorithms that can be implemented in software.
3. Consensus methods for phylogenetic networks. Statistical tools that quantify uncertainty and robustly summarize combinatorial features of phylogenetic networks are needed to draw biologically meaningful conclusions from collections of inferred networks. We will identify features of phylogenetic networks that are best suited to develop consensus methods and analyze how they can be inferred from data.

The seminar will be split between research talks from both research communities and breakout sessions in which small groups of participants work on specific problems to initiate new collaborations.

Creative Commons BY 4.0

Laura Kubatko, Simone Linz, Leo van Iersel, and Kristina Wicke