Systematic Biology最新文献

The recent rapid radiation of Tillandsia subgenus Tillandsia (Bromeliaceae) provides an attractive system to study the drivers and constraints of species diversification. This species-rich Neotropical monocot clade includes predominantly epiphytic species displaying vast phenotypic diversity. Recent in-depth phylogenomic work revealed that the subgenus originated within the last 7 myr, with one major expansion from South into Central America within the last 5 myr. However, disagreements between phylogenies and lack of resolution at shallow nodes suggest that hybridization may have occurred throughout the radiation, together with frequent incomplete lineage sorting and rapid gene family evolution. We used whole-genome resequencing data to explore the evolutionary history of representative ingroup species employing both tree-based and network approaches. Our results indicate that lineage co-occurrence does not predict relatedness and confirm significant deviations from a tree-like structure, coupled with pervasive gene-tree discordance. Focusing on hybridization, ABBA-BABA and related statistics were used to infer the rates and relative timing of introgression, whereas topology weighting uncovered high heterogeneity of the phylogenetic signal along the genome. High rates of hybridization within and among subclades suggest that, contrary to previous hypotheses, the expansion of subgenus Tillandsia into Central America proceeded through several dispersal events, punctuated by episodes of diversification and gene flow. Network analysis revealed reticulation as a plausible propeller during radiation and establishment across different ecological niches. This work contributes a plant example of prevalent hybridization during rapid species diversification, supporting the hypothesis that interspecific gene flow facilitates explosive diversification.

Bayesian inference has predominantly relied on the Markov chain Monte Carlo (MCMC) algorithm for many years. However, MCMC is computationally laborious, especially for complex phylogenetic models of time trees. This bottleneck has led to the search for alternatives, such as variational Bayes, which can scale better to large data sets. In this paper, we introduce torchtree, a framework written in Python that allows developers to easily implement rich phylogenetic models and algorithms using a fixed tree topology. One can either use automatic differentiation or leverage torchtree's plug-in system to compute gradients analytically for model components for which automatic differentiation is slow. We demonstrate that the torchtree variational inference framework performs similarly to BEAST in terms of speed, and delivers promising approximation results, though accuracy varies across scenarios. Furthermore, we explore the use of the forward Kullback-Leibler (KL) divergence as an optimizing criterion for variational inference, which can handle discontinuous and nondifferentiable models. Our experiments show that inference using the forward KL divergence is frequently faster per iteration compared with the evidence lower bound (ELBO) criterion, although the ELBO-based inference may converge faster in some cases. Overall, torchtree provides a flexible and efficient framework for phylogenetic model development and inference using PyTorch.

A somewhat personal account of the development and acceptance of numerical taxonomic methods during the early years of the journal Systematic Zoology. Includes a few perspectives on the changes in taxonomy and the journal after 75 years.

Unrooted phylogenetic networks are commonly used to represent evolutionary data in the presence of incompatibilities. Although rooted phylogenetic networks offer a more explicit framework for depicting evolutionary histories involving reticulate events, they are reported less frequently, probably due to a lack of tools that are as easily applicable as those for unrooted networks. Here, we introduce PhyloFusion, a fast and user-friendly method for constructing rooted phylogenetic networks from sets of rooted phylogenetic trees. The resulting networks have the tree-child property. The algorithm accommodates trees with unresolved nodes-often resulting from the contraction of low-support edges-as well as some degree of missing taxa. We demonstrate its application to the analysis of functionally related gene groups and show that it can efficiently handle data sets comprising tens of trees or hundreds of taxa. An open source implementation of PhyloFusion is available as part of the SplitsTree app: https://www.github.com/husonlab/splitstree6. All data available here: https://doi.org/10.5061/dryad.k3j9kd5h5.

Missing data is a long-standing issue in phylogenetic inference, which often results in high levels of taxonomic instability, obscuring otherwise well-supported relationships. Multiple approaches have been developed to deal with the negative effects of ineffective overlap on tree resolution, often by identifying taxa for removal. Here, we repurpose a heuristic method developed to identify unstable taxa in morphological data matrices, concatabominations, and combine it with a novel gene-tree jackknifing on matrix representation of trees to identify candidates for targeted sequencing. Using a multilocus caecilian data set, we illustrate the method's capacity to identify candidate taxa and loci for additional sequencing, compare the results with those of the mathematics-based gene sampling sufficiency approach, and explore the terrace space associated with the multilocus data set. We show that our approach yields tractable numbers of loci/taxa for targeted sequencing that successfully mitigate topological instability due to ineffective overlap, even when modest amounts of data are added.

Reconstructing phylogeny from morphological data remains mired in investigator biases, including subjective inclusion and discretisation of phenotypic variation. Geometric morphometrics and multivariate statistical analyses provide an alternative array of tools for studying variation in morphological traits. However, direct analysis of landmark data is often unreliable for phylogeny reconstruction. Morphological variation is typically highly correlated among nearby landmarks and may evolve saltationally between adaptive peaks instead of gradually, thereby violating the assumptions of typical continuous models. To address these concerns, we developed an approach to more objectively discretise morphometric data and applied it to 3D surface scans of mandibles and postcranial elements of Macropodiformes (kangaroos, bettongs and rat-kangaroos). The scanned elements were partitioned into sets of locally co-varying landmarks which approximate functional units. These subregions were discretised into "atomised" characters using novel approaches to combine the objectivity of continuous shape variation for delineating discrete states with the model flexibility offered for multistate and binary characters. This allows us to (1) potentially reduce the influence of non-independence among neighbouring landmarks, (2) accommodate multimodal variation from saltational evolution, (3) accommodate missing data, such as from fragmentary fossils, and (4) promote tree-search efficiency. We built discrete morphological character matrices using three alternative approaches: commonly used clustering algorithms (UPGMA, k-means, k-medoids, Gaussian mixture modelling), a minimum evolution branch length criterion, and a tree sampling procedure. Our phylogenetic analyses with these novel matrices generally succeeded in recovering genera and several deep-level macropodiform clades, but failed to accurately reconstruct intergeneric relationships within the rapid diversification of the macropodine sub-family; those relationships were also not recovered with continuous morphological data or traditionally discretised characters and are the most poorly resolved with DNA data. On balance, our atomised characters, which derive from only mandibular and three postcranial elements, show promise for improving objectivity, accuracy and clocklikeness in morphological phylogenetics and provide pathways for accommodating correlated homoplasy and for more accurately estimating rates of morphological evolution, and thereby better integrating phenotypic and genomic data for phylogenetic inference.