Unsupervised machine learning for species delimitation, integrative taxonomy, and biodiversity conservation.

Integrative taxonomy, combining data from multiple axes of biologically relevant variation, is a major goal of systematics. Ideally, such taxonomies will derive from similarly integrative species-delimitation analyses. Yet, most current methods rely solely or primarily on molecular data, with other layers often incorporated only in a post hoc qualitative or comparative manner. A major limitation is the difficulty of devising quantitative parametric models linking different datasets in a unified ecological and evolutionary framework. Machine Learning (ML) methods offer flexibility in this arena by easily learning high-dimensional associations between observations (e.g., individual specimens) across a wide array of input features (e.g., genetics, geography, environment, and phenotype) to delimit statistically meaningful clusters. Here, I implement an unsupervised method using Self-Organizing (or "Kohonen") Maps (SOMs) for such purposes. Recent extensions called "SuperSOMs" can integrate multiple layers, each of which exerts independent influence on a two-dimensional output grid via empirically estimated weights. The grid cells are then delimited into K distinct units that can be interpreted as species or other entities. I show empirical examples in salamanders (Desmognathus) and snakes (Storeria) with layers representing alleles, space, climate, and traits. Simulations reveal that the SuperSOM approach can detect K = 1, tends not to over-split, reflects contributions from all layers, and limits large layers (e.g., genetic matrices) from overwhelming other datasets, desirable properties addressing major concerns from previous studies. Finally, I suggest that these and similar methods could integrate conservation-relevant layers such as population trends and human encroachment to delimit management units from an explicitly quantitative framework grounded in the ecology and evolution of species limits and boundaries.

An evaluation of fossil tip-dating versus node-age calibrations in tetraodontiform fishes (Teleostei: Percomorphaceae).

Time-calibrated phylogenies based on molecular data provide a framework for comparative studies. Calibration methods to combine fossil information with molecular phylogenies are, however, under active development, often generating disagreement about the best way to incorporate paleontological data into these analyses. This study provides an empirical comparison of the most widely used approach based on node-dating priors for relaxed clocks implemented in the programs BEAST and MrBayes, with two recently proposed improvements: one using a new fossilized birth-death process model for node dating (implemented in the program DPPDiv), and the other using a total-evidence or tip-dating method (implemented in MrBayes and BEAST). These methods are applied herein to tetraodontiform fishes, a diverse group of living and extinct taxa that features one of the most extensive fossil records among teleosts. Previous estimates of time-calibrated phylogenies of tetraodontiforms using node-dating methods reported disparate estimates for their age of origin, ranging from the late Jurassic to the early Paleocene (ca. 150-59Ma). We analyzed a comprehensive dataset with 16 loci and 210 morphological characters, including 131 taxa (95 extant and 36 fossil species) representing all families of fossil and extant tetraodontiforms, under different molecular clock calibration approaches. Results from node-dating methods produced consistently younger ages than the tip-dating approaches. The older ages inferred by tip dating imply an unlikely early-late Jurassic (ca. 185-119Ma) origin for this order and the existence of extended ghost lineages in their fossil record. Node-based methods, by contrast, produce time estimates that are more consistent with the stratigraphic record, suggesting a late Cretaceous (ca. 86-96Ma) origin. We show that the precision of clade age estimates using tip dating increases with the number of fossils analyzed and with the proximity of fossil taxa to the node under assessment. This study suggests that current implementations of tip dating may overestimate ages of divergence in calibrated phylogenies. It also provides a comprehensive phylogenetic framework for tetraodontiform systematics and future comparative studies.

Apparent signal of competition limiting diversification after ecological transitions from marine to freshwater habitats.

Adaptive radiations are typically triggered when a lineage encounters a significant range of open niche space (ecological opportunity), stemming from colonisation of new areas, extinction of competitors or key innovations. The most well-known of these is the colonisation of new areas, through either dispersal into new regions or the invasion of novel ecological regimes. One aspect of ecological opportunity that has rarely been studied, however, is the extent to which pre-existent competitors act to limit diversification in newly colonised adaptive zones. Herein, we show that in multiple geographically independent invasions of freshwaters by marine Sea Catfishes (Ariidae), rates of both morphological disparification and lineage diversification are inversely related to the presence and diversity of other freshwater fish lineages. Only in one region (Australia-New Guinea) with an otherwise depauperate freshwater fauna, has an ariid invasion gained any substantial traction. This is true at both regional and community scales, suggesting that competitive constraints may be an important factor regulating adaptive radiation.

Lineage diversification in a widespread species: roles for niche divergence and conservatism in the common kingsnake, Lampropeltis getula.

Niche conservatism and niche divergence are both important ecological mechanisms associated with promoting allopatric speciation across geographical barriers. However, the potential for variable responses in widely distributed organisms has not been fully investigated. For allopatric sister lineages, three patterns for the interaction of ecological niche preference and geographical barriers are possible: (i) niche conservatism at a physical barrier; (ii) niche divergence at a physical barrier; and (iii) niche divergence in the absence of a physical barrier. We test for the presence of these patterns in a transcontinentally distributed snake species, the common kingsnake (Lampropeltis getula), to determine the relative frequency of niche conservatism or divergence in a single species complex inhabiting multiple distinct ecoregions. We infer the phylogeographic structure of the kingsnake using a range-wide data set sampled for the mitochondrial gene cytochrome b. We use coalescent simulation methods to test for the presence of structured lineage formation vs. fragmentation of a widespread ancestor. Finally, we use statistical techniques for creating and evaluating ecological niche models to test for conservatism of ecological niche preferences. Significant geographical structure is present in the kingsnake, for which coalescent tests indicate structured population division. Surprisingly, we find evidence for all three patterns of conservatism and divergence. This suggests that ecological niche preferences may be labile on recent phylogenetic timescales, and that lineage formation in widespread species can result from an interaction between inertial tendencies of niche conservatism and natural selection on populations in ecologically divergent habitats.

Phylogeography across a continent: the evolutionary and demographic history of the North American racer (Serpentes: Colubridae: Coluber constrictor).

Most phylogeographic studies examine organisms that do not have transcontinental distributions and therefore the genetic and temporal effects of barriers across an entire continent cannot be assessed with respect to a single species. We examined the phylogeographic structure, lineage age, and historical demography using sequences from the mtDNA cytochrome b gene of the widespread North American racer (Coluber constrictor), one of the few abundant transcontinental snakes that occurs throughout many diverse biomes. Our results indicate that this complex is comprised of six lineages differing greatly in geographic extent, with the largest (a central US clade) being approximately 26 times greater than the smallest (a lineage restricted to the Florida Panhandle and nearby portions of adjacent States). Most of the six lineages appear to be separated at previously identified genetic barriers for several vertebrates with similar ranges. Lineage diversification in this species began in the late Miocene, separating populations in the Florida Peninsula from the remainder of the US. Diversification of lineages continued throughout the Pliocene and early Pleistocene. Four of the six lineages occur east of the Mississippi River, with only two distinctly young ( approximately 1.5 mya) lineages found west of the Mississippi River (one occurs west of Continental Divide). All methods of demographic inference, including the mismatch distribution, Fu and Li's D and Tajima's D , and Bayesian skyline plots revealed population expansion occurring in the mid-to-late Pleistocene for every lineage, regardless of size or proximity to formerly glaciated areas. Population expansion for lineages found east of the Mississippi River occurred earlier and was much greater than those found west of the River.