Bayesian Estimation Of Species Trees Research Articles

An improved phylogeny of the Andean tit-tyrants (Aves, Tyrannidae): More characters trump sophisticated analyses

The phylogeny of the flycatcher genus Anairetes was previously inferred using short fragments of mitochondrial DNA and parsimony and distance-based methods. The resulting topology spurred taxonomic revision and influenced understanding of Andean biogeography. More than a decade later, we revisit the phylogeny of Anairetes tit-tyrants using more mtDNA characters, seven unlinked loci (three mitochondrial genes, six nuclear loci), more closely related outgroup taxa, partitioned Bayesian analyses, and two coalescent species-tree approaches (Bayesian estimation of species trees, BEST; Bayesian evolutionary analysis by sampling trees, *BEAST). Of these improvements in data and analyses, the fourfold increase in mtDNA characters was both necessary and sufficient to incur a major shift in the topology and near-complete resolution. The species-tree analyses, while theoretically preferable to concatenation or single gene approaches, yielded topologies that were compatible with mtDNA but with weaker statistical resolution at nodes. The previous results that had led to taxonomic and biogeographic reappraisal were refuted, and the current results support the resurrection of the genus Uromyias as the sister clade to Anairetes. The sister relationship between these two genera corresponds to an ecological dichotomy between a depauperate humid cloud forest clade and a diverse dry-tolerant clade that has diversified along the latitudinal axis of the Andes. The species-tree results and the concatenation results each reaffirm the primacy of mtDNA to provide phylogenetic signal for avian phylogenies at the species and subspecies level. This is due in part to the abundance of informative characters in mtDNA, and in part to its lower effective population size that causes it to more faithfully track the species tree.

Phylogenetics and genetic diversity of the Cotesia flavipes complex of parasitoid wasps (Hymenoptera: Braconidae), biological control agents of lepidopteran stemborers

The Cotesia flavipes complex of parasitoid wasps (Hymenoptera: Braconidae) are economically important for the biological control of lepidopteran stemboring pests associated with gramineous crops. Some members of the complex successfully parasitize numerous stemborer pest species, however certain geographic populations have demonstrated variation in the range of hosts that they parasitize. In addition, the morphology of the complex is highly conserved and considerable confusion surrounds the identity of species and host-associated biotypes. We generated nucleotide sequence data for two mtDNA genes (COI, 16S) and three anonymous nuclear loci (CfBN, CfCN, CfEN) for the C. flavipes complex. To analyze genetic variation and relationships among populations we used (1) concatenated mtDNA and nDNA data, (2) a nDNA multilocus network approach, and (3) two species tree inference methods, i.e. Bayesian estimation of species trees (BEST) and Bayesian inference of species trees from multilocus data with *BEAST. All phylogenetic analyses provide strong support for monophyly of the complex and the presence of at least four species, C. chilonis (from China and Japan), C. sesamiae (from Africa), C. flavipes (originating from the Indo-Asia region but introduced into Africa and the New World), and C. nonagriae (from Australia and Papua New Guinea). Haplotype diversity of geographic populations relates to historical biogeographic barriers and biological control introductions, and reflects previous reports of ecological variation in these species. Strong discordance was found between the mitochondrial and nuclear markers in the Papua New Guinea haplotypes, which may be an outcome of hybridization and introgression of C. flavipes and C. nonagriae. The position of Cotesia flavipes from Japan was not well supported in any analysis and was the sister taxon to C. nonagriae (mtDNA, *BEAST), C. flavipes (nDNA) or C. flavipes+C. nonagriae (BEST) and, may represent a cryptic species. The concatenated five gene phylogenetic analyses did not support the overall separation and monophyly of clades associated with different host species, although some clades did show specific host associations, possibly due to localized host availability, rather than host specificity. Our results provide a framework for assessing whether distinct lineages represent cryptic species, and for examining parasitoid–host evolution and compatibility more generally. Given the limitations of morphological based identification for members of this complex, molecular identification is recommended prior to any biological control introductions.

MrBayes 3.2: efficient Bayesian phylogenetic inference and model choice across a large model space.

Since its introduction in 2001, MrBayes has grown in popularity as a software package for Bayesian phylogenetic inference using Markov chain Monte Carlo (MCMC) methods. With this note, we announce the release of version 3.2, a major upgrade to the latest official release presented in 2003. The new version provides convergence diagnostics and allows multiple analyses to be run in parallel with convergence progress monitored on the fly. The introduction of new proposals and automatic optimization of tuning parameters has improved convergence for many problems. The new version also sports significantly faster likelihood calculations through streaming single-instruction-multiple-data extensions (SSE) and support of the BEAGLE library, allowing likelihood calculations to be delegated to graphics processing units (GPUs) on compatible hardware. Speedup factors range from around 2 with SSE code to more than 50 with BEAGLE for codon problems. Checkpointing across all models allows long runs to be completed even when an analysis is prematurely terminated. New models include relaxed clocks, dating, model averaging across time-reversible substitution models, and support for hard, negative, and partial (backbone) tree constraints. Inference of species trees from gene trees is supported by full incorporation of the Bayesian estimation of species trees (BEST) algorithms. Marginal model likelihoods for Bayes factor tests can be estimated accurately across the entire model space using the stepping stone method. The new version provides more output options than previously, including samples of ancestral states, site rates, site dN/dS rations, branch rates, and node dates. A wide range of statistics on tree parameters can also be output for visualization in FigTree and compatible software.

A 10‐gene phylogeny of Solanum section Herpystichum (Solanaceae) and a comparison of phylogenetic methods

Solanum section Herpystichum is a lineage that comprises both widespread and very narrowly distributed species. This study investigates the phylogenetic relationships of sect. Herpystichum and evaluates several phylogenetic methods for analysis of multiple sequences. Sequence data from seven nuclear (ITS, GBSSI, and five COSII) and three plastid (psbA-trnH, trnT-trnF, and trnS-trnG) regions were concatenated and analyzed under maximum parsimony and Bayesian criteria. In addition, we used two analytical methods that take into account differences in topologies resulting from the analyses of the individual markers: Bayesian Estimation of Species Trees (BEST) and supertree analysis. The monophyletic Solanum sect. Herpystichum was resolved with moderate support in the concatenated maximum parsimony and Bayesian analyses and the supertree analysis, and relationships within the section were well-resolved and strongly supported. The BEST topology, however, was poorly resolved. Also, because of how BEST deals with missing sequences, >25% of our accessions, including two species, had to be excluded from the analyses. Our results indicate a progenitor-descendent relationship with two species nested within the widespread S. evolvulifolium. Analytical methods that consider individual topologies are important for studies based on multiple molecular markers. On the basis of analyses in this study, BEST had the serious shortcoming that taxa with missing sequences must be removed from the analysis or they can produce spurious topologies. Supertree analysis provided a good alternative for our data by allowing the inclusion of all 10 species of sect. Herpystichum.

43 genes support the lungfish-coelacanth grouping related to the closest living relative of tetrapods with the Bayesian method under the coalescence model

BackgroundSince the discovery of the "living fossil" in 1938, the coelacanth (Latimeria chalumnae) has generally been considered to be the closest living relative of the land vertebrates, and this is still the prevailing opinion in most general biology textbooks. However, the origin of tetrapods has not been resolved for decades. Three principal hypotheses (lungfish-tetrapod, coelacanth-tetrapod, or lungfish-coelacanth sister group) have been proposed.FindingsWe used the Bayesian method under the coalescence model with the latest published program (Bayesian Estimation of Species Trees, or BEST) to perform a phylogenetic analysis for seven relevant taxa and 43 nuclear protein-coding genes with the jackknife method for taxon sub-sampling. The lungfish-coelacanth sister group was consistently reconstructed with the Bayesian method under the coalescence model in 17 out of 21 taxon sets with a Bayesian posterior probability as high as 99%. Lungfish-tetrapod was only inferred from BCLS and BACLS. Neither coelacanth-tetrapod nor lungfish-coelacanth-tetrapod was recovered out of all 21 taxon sets.ConclusionsOur results provide strong evidence in favor of accepting the hypothesis that lungfishes and coelacanths form a monophyletic sister-group that is the closest living relative of tetrapods. This clade was supported by high Bayesian posterior probabilities of the branch (a lungfish-coelacanth clade) and high taxon jackknife supports.

Comparing Two Bayesian Methods for Gene Tree/Species Tree Reconstruction: Simulations with Incomplete Lineage Sorting and Horizontal Gene Transfer

With the increasing interest in recognizing the discordance between gene genealogies, various gene tree/species tree reconciliation methods have been developed. We present here the first attempt to assess and compare two such Bayesian methods, Bayesian estimation of species trees (BEST) and BUCKy (Bayesian untangling of concordance knots), in the presence of several known processes of gene tree discordance. DNA alignments were simulated under the influence of incomplete lineage sorting (ILS) and of horizontal gene transfer (HGT). BEST and BUCKy both account for uncertainty in gene tree estimation but differ substantially in their assumptions of what caused gene tree discordance. BEST estimates a species tree using the coalescent model, assuming that all gene tree discordance is due to ILS. BUCKy does not assume any specific biological process of gene tree discordance through the use of a nonparametric clustering of concordant genes. BUCKy estimates the concordance factor (CF) of a clade, which is defined as the proportion of genes that truly have the clade in their trees. The estimated concordance tree is then built from clades with the highest estimated CFs. Because of their different assumptions, it was expected that BEST would perform better in the presence of ILS and that BUCKy would perform better in the presence of HGT. As expected, the species tree was more accurately reconstructed by BUCKy in the presence of HGT, when the HGT events were unevenly placed across the species tree. BUCKy and BEST performed similarly in most other cases, including in the presence of strong ILS and of HGT events that were evenly placed across the tree. However, BUCKy was shown to underestimate the uncertainty in CF estimation, with short credibility intervals. Despite this, the discordance pattern estimated by BUCKy could be compared with the signature of ILS. The resulting test for the adequacy of the coalescent model proved to have low Type I error. It was powerful when HGT was the major source of discordance and when HGT events were unevenly placed across the species tree.

Multilocus analysis of a taxonomically densely sampled dataset reveal extensive non-monophyly in the avian family Locustellidae

The phylogeny of most of the species in the avian passerine family Locustellidae is inferred using a Bayesian species tree approach (Bayesian Estimation of Species Trees, BEST), as well as a traditional Bayesian gene tree method (MrBayes), based on a dataset comprising one mitochondrial and four nuclear loci. The trees inferred by the different methods agree fairly well in topology, although in a few cases there are marked differences. Some of these discrepancies might be due to convergence problems for BEST (despite up to 1×109 iterations). The phylogeny strongly disagrees with the current taxonomy at the generic level, and we propose a revised classification that recognizes four instead of seven genera. These results emphasize the well known but still often neglected problem of basing classifications on non-cladistic evaluations of morphological characters. An analysis of an extended mitochondrial dataset with multiple individuals from most species, including many subspecies, suggest that several taxa presently treated as subspecies or as monotypic species as well as a few taxa recognized as separate species are in need of further taxonomic work.

Phylogenetic relationships in the spoon tarsus subgroup of Hawaiian drosophila: Conflict and concordance between gene trees

The Hawaiian Drosophilidae contains approximately 1000 species, placed in species groups and subgroups based largely on secondary sexual modifications to wings, forelegs and mouthparts. Members of the spoon tarsus subgroup possess a cup-shaped structure on the foretarsi of males. Eight of the twelve species in this subgroup are found only on the Big Island of Hawaii, suggesting that they have diverged within the past 600,000years. This rapid diversification has made determining the relationships within this group difficult to infer. We use 13 genes, including nine rapidly evolving nuclear loci, to estimate relationships within the spoon tarsus species, as well as to test the monophyly of this subgroup. A variety of analytical approaches are used, including individual and concatenated analyses, Bayesian estimation of species trees and Bayesian untangling of concordance knots. We find widespread agreement between phylogenetic estimates derived from different methods, although some incongruence is present. Notably, our analyses suggest that the spoon tarsus subgroup, as currently defined, is not monophyletic.

Gene trees, species and species trees in the Ctenosaura palearis clade

The growing use of molecular systematics in conservation has increased the importance of accurate resolution of taxonomic units and relationships. DNA data relate most directly to genealogies, which need not have perfect relationships with species limits and phylogenies. We used a multilocus gene tree approach to elucidate the relationships between four endangered Central American iguanas. We found support for the proposition that the described species taxa correspond to distinct evolutionary lineages warranting individual protection. We combined gene trees to estimate a phylogeny using Bayesian Estimation of Species Trees (BEST), minimizing deep coalescence, Species Trees from Average Ranks (STAR), and traditional concatenation. The estimate from concatenation conflicted with the other methods, likely owing to the disproportionate effect of mtDNA on concatenated analyses. This illustrates the importance of appropriate treatment of multilocus sequence data in phylogenetics. Our results indicate that these species have gone through recent and rapid speciation, resulting in four closely related narrow-range endemics.

Gene Trees versus Species Trees: Reassessing Life-History Evolution in a Freshwater Fish Radiation

Mechanisms of speciation are best understood in the context of phylogenetic relationships and as such have often been inferred from single gene trees, typically those derived from mitochondrial DNA (mtDNA) markers. Recent studies, however, have noted the potential for phylogenetic discordance between gene trees and underlying species trees (e.g., due to stochastic lineage sorting, introgression, or selection). Here, we employ a variety of nuclear DNA loci to reassess evolutionary relationships within a recent freshwater fish radiation to reappraise modes of speciation. New Zealand's freshwater-limited Galaxias vulgaris complex is thought to have evolved from G. brevipinnis, a widespread migratory species that retains a plesiomorphic marine juvenile phase. A well-resolved tree, based on four mtDNA regions, previously suggested that marine migratory ability has been lost on 3 independent occasions in the evolution of this species flock (assuming that loss of diadromy is irreversible). Here, we use pseudogene (galaxiid Numt: 1801 bp), intron (S: 903 bp), and exon (RAG-1: 1427 bp) markers, together with mtDNA, to reevaluate this hypothesis of parallel evolution. Interestingly, partitioned Bayesian analysis of concatenated nuclear sequences (3141 bp) and concatenated nuclear and mtDNA (4770 bp) both recover phylogenies implying a single loss of diadromy, not three parallel losses as previously inferred from mtDNA alone. This phylogenetic result is reinforced by a multilocus analysis performed using Bayesian estimation of species trees (BEST) software that estimates the posterior distribution of species trees under a coalescent model. We discuss factors that might explain the apparently misleading phylogenetic inferences generated by mtDNA.

Supermatrix and species tree methods resolve phylogenetic relationships within the big cats, Panthera (Carnivora: Felidae)

The pantherine lineage of cats diverged from the remainder of modern Felidae less than 11 million years ago and consists of the five big cats of the genus Panthera, the lion, tiger, jaguar, leopard, and snow leopard, as well as the closely related clouded leopard. A significant problem exists with respect to the precise phylogeny of these highly threatened great cats. Despite multiple publications on the subject, no two molecular studies have reconstructed Panthera with the same topology. These evolutionary relationships remain unresolved partially due to the recent and rapid radiation of pantherines in the Pliocene, individual speciation events occurring within less than 1 million years, and probable introgression between lineages following their divergence. We provide an alternative, highly supported interpretation of the evolutionary history of the pantherine lineage using novel and published DNA sequence data from the autosomes, both sex chromosomes and the mitochondrial genome. New sequences were generated for 39 single-copy regions of the felid Y chromosome, as well as four mitochondrial and four autosomal gene segments, totaling 28.7 kb. Phylogenetic analysis of these new data, combined with all published data in GenBank, highlighted the prevalence of phylogenetic disparities stemming either from the amplification of a mitochondrial to nuclear translocation event (numt), or errors in species identification. Our 47.6 kb combined dataset was analyzed as a supermatrix and with respect to individual partitions using maximum likelihood and Bayesian phylogenetic inference, in conjunction with Bayesian Estimation of Species Trees (BEST) which accounts for heterogeneous gene histories. Our results yield a robust consensus topology supporting the monophyly of lion and leopard, with jaguar sister to these species, as well as a sister species relationship of tiger and snow leopard. These results highlight new avenues for the study of speciation genomics and understanding the historical events surrounding the origin of the members of this lineage.

Species Tree Discordance Traces to Phylogeographic Clade Boundaries in North American Fence Lizards (Sceloporus)

I investigated the impacts of phylogeographic sampling decisions on species tree estimation in the Sceloporus undulatus species group, a recent radiation of small, insectivorous lizards connected by parapatric and peripatric distribution across North America, using a variety of species tree inference methods (Bayesian estimation of species trees, Bayesian untangling of concordance knots, and minimize deep coalescences). Phylogenetic analyses of 16 specimens representing 4 putative species within S. "undulatus" using complete (8 loci, >5.5 kb) and incomplete (29 loci, >23.6 kb) nuclear data sets result in species trees that share features with the mitochondrial DNA (mtDNA) genealogy at the phylogeographic level but provide new insights into the evolutionary history of the species group. The concatenated nuclear data and mtDNA data both recover 4 major clades connecting populations across North America; however, instances of discordance are localized at the contact zones between adjacent phylogeographic groups. A random sub-sampling experiment designed to vary the phylogeographic samples included across hundreds of replicate species tree inferences suggests that inaccurate species assignments can result in inferred phylogenetic relationships that are dependent upon which particular populations are used as exemplars to represent species and can lead to increased estimates of effective population size. For the phylogeographic data presented here, reassigning specimens with introgressed mtDNA genomes to their prospective species, or excluding them from the analysis altogether, produces species tree topologies that are distinctly different from analyses that utilize mtDNA-based species assignments. Evolutionary biologists working at the interface of phylogeography and phylogenetics are likely to encounter multiple processes influencing gene trees congruence, which increases the relevance of estimating species trees with multilocus nuclear data and models that accommodate deep coalescence.

Species Trees from Highly Incongruent Gene Trees in Rice

Several methods have recently been developed to infer multilocus phylogenies by incorporating information from topological incongruence of the individual genes. In this study, we investigate 2 such methods, Bayesian concordance analysis and Bayesian estimation of species trees. Our test data are a collection of genes from cultivated rice (genus Oryza) and the most closely related wild species, generated using a high-throughput sequencing protocol and bioinformatics pipeline. Trees inferred from independent genes display levels of topological incongruence that far exceed that seen in previous data sets analyzed with these species tree methods. We identify differences in phylogenetic results between inference methods that incorporate gene tree incongruence. Finally, we discuss the challenges of scaling these analyses for data sets with thousands of gene trees and extensive levels of missing data.

The utility of nuclear conserved ortholog set II (COSII) genomic regions for species-level phylogenetic inference in Lycium (Solanaceae)

The identification of genomic regions with sufficient variation to elucidate fine-scale relationships among closely related species is a major goal of phylogenetic systematics. However, the accumulation of such multi-locus data sets brings its own challenges, given that gene trees do not necessarily represent the true species tree. Using genomic tools developed for Solanum (Solanaceae), we have evaluated the utility of nuclear conserved ortholog set II (COSII) regions for phylogenetic inference in tribe Lycieae (Solanaceae). Five COSII regions, with intronic contents ranging from 68% to 91%, were sequenced in 10 species. Their phylogenetic utility was assessed and compared with data from more commonly used nuclear (GBSSI, nrITS) and cpDNA spacer data. We compared the effectiveness of a traditional total evidence concatenation approach versus the recently developed Bayesian estimation of species trees (BEST) method to infer species trees given multiple independent gene trees. All of the sampled COSII regions had high numbers of parsimony-informative (PI) characters, and two of the COSII regions had more PI characters than the GBSSI, ITS, and cpDNA spacer data sets combined. COSII markers are a promising new tool for phylogenetic inference in Solanaceae, and should be explored in related groups. Both the concatenation and BEST approaches yielded similar topologies; however, when multiple individuals with polyphyletic alleles were included, BEST was clearly the more robust approach for inferring species trees in the presence of gene tree incongruence.

Comparison of Methods for Species-Tree Inference in the Sawfly Genus Neodiprion (Hymenoptera: Diprionidae)

Conifer-feeding sawflies in the genus Neodiprion provide an excellent opportunity to investigate the origin and maintenance of barriers to reproduction, but obtaining a phylogenetic estimate for comparative studies of Neodiprion speciation has proved difficult. Specifically, nonmonophyly within and discordance between individual gene trees, both of which are common in groups that diverged recently and/or rapidly, make it impossible to infer a species tree using methods that are designed to estimate gene trees. Therefore, in this study, we estimate relationships between members of the lecontei species group using four approaches that are intended to estimate species, not gene, trees: (1) minimize deep coalescences (MDC), (2) shallowest divergences (SD), (3) Bayesian estimation of species trees (BEST), and (4) a novel approach that combines concatenation with monophyly constraints (CMC). Multiple populations are sampled for most species and all four methods incorporate this intraspecific variation into estimates of interspecific relationships. We investigate the sensitivity of each method to taxonomic sampling, and, for the BEST method, we assess the impact of prior choice on species-tree inference. We also compare species-tree estimates to one another and to a morphologically based hypothesis to identify clades that are supported by multiple analyses and lines of evidence. We find that both taxonomic sampling and method choice impact species-tree estimates and that, for these data, the BEST method is strongly influenced by Theta and branch-length priors. We also find that the CMC method is the least sensitive to taxonomic sampling. Finally, although interspecific genetic variation is low due to the recent divergence of the lecontei group, our results to date suggest that incomplete lineage sorting and interspecific gene flow are the main factors complicating species-tree inference in Neodiprion. Based on these analyses, we propose a phylogenetic hypothesis for the lecontei group. Finally, our results suggest that, even for very challenging groups like Neodiprion, an underlying species-tree signal can be extracted from multi-locus data as long as intraspecific variation is adequately sampled and methods that focus on the estimation of species trees are used.

BEST: Bayesian estimation of species trees under the coalescent model

BEST implements a Bayesian hierarchical model to jointly estimate gene trees and the species tree from multilocus sequences. It provides a new option for estimating species phylogenies within the popular Bayesian phylogenetic program MrBayes. The technique of simulated annealing is adopted along with Metropolis coupling as performed in MrBayes to improve the convergence rate of the Markov Chain Monte Carlo algorithm. http://www.stat.osu.edu/~dkp/BEST.