Jeremy M. Brown

The Effect of Ambiguous Data on Phylogenetic Estimates Obtained by Maximum Likelihood and Bayesian Inference

Abstract Although an increasing number of phylogenetic data sets are incomplete, the effect of ambiguous data on phylogenetic accuracy is not well understood. We use 4-taxon simulations to study the effects of ambiguous data (i.e., missing characters or gaps) in maximum likelihood (ML) and Bayesian frameworks. By introducing ambiguous data in a way that removes confounding factors, we provide the first clear understanding of 1 mechanism by which ambiguous data can mislead phylogenetic analyses. We find that in both ML and Bayesian frameworks, among-site rate variation can interact with ambiguous data to produce misleading estimates of topology and branch lengths. Furthermore, within a Bayesian framework, priors on branch lengths and rate heterogeneity parameters can exacerbate the effects of ambiguous data, resulting in strongly misleading bipartition posterior probabilities. The magnitude and direction of the ambiguous data bias are a function of the number and taxonomic distribution of ambiguous characters, the strength of topological support, and whether or not the model is correctly specified. The results of this study have major implications for all analyses that rely on accurate estimates of topology or branch lengths, including divergence time estimation, ancestral state reconstruction, tree-dependent comparative methods, rate variation analysis, phylogenetic hypothesis testing, and phylogeographic analysis.

The Importance of Data Partitioning and the Utility of Bayes Factors in Bayesian Phylogenetics

As larger, more complex data sets are being used to infer phylogenies, accuracy of these phylogenies increasingly requires models of evolution that accommodate heterogeneity in the processes of molecular evolution. We investigated the effect of improper data partitioning on phylogenetic accuracy, as well as the type I error rate and sensitivity of Bayes factors, a commonly used method for choosing among different partitioning strategies in Bayesian analyses. We also used Bayes factors to test empirical data for the need to divide data in a manner that has no expected biological meaning. Posterior probability estimates are misleading when an incorrect partitioning strategy is assumed. The error was greatest when the assumed model was underpartitioned. These results suggest that model partitioning is important for large data sets. Bayes factors performed well, giving a 5% type I error rate, which is remarkably consistent with standard frequentist hypothesis tests. The sensitivity of Bayes factors was found to be quite high when the across-class model heterogeneity reflected that of empirical data. These results suggest that Bayes factors represent a robust method of choosing among partitioning strategies. Lastly, results of tests for the inclusion of unexpected divisions in empirical data mirrored the simulation results, although the outcome of such tests is highly dependent on accounting for rate variation among classes. We conclude by discussing other approaches for partitioning data, as well as other applications of Bayes factors.

When trees grow too long: investigating the causes of highly inaccurate bayesian branch-length estimates.

A surprising number of recent Bayesian phylogenetic analyses contain branch-length estimates that are several orders of magnitude longer than corresponding maximum-likelihood estimates. The levels of divergence implied by such branch lengths are unreasonable for studies using biological data and are known to be false for studies using simulated data. We conducted additional Bayesian analyses and studied approximate-posterior surfaces to investigate the causes underlying these large errors. We manipulated the starting parameter values of the Markov chain Monte Carlo (MCMC) analyses, the moves used by the MCMC analyses, and the prior-probability distribution on branch lengths. We demonstrate that inaccurate branch-length estimates result from either 1) poor mixing of MCMC chains or 2) posterior distributions with excessive weight at long tree lengths. Both effects are caused by a rapid increase in the volume of branch-length space as branches become longer. In the former case, both an MCMC move that scales all branch lengths in the tree simultaneously and the use of overdispersed starting branch lengths allow the chain to accurately sample the posterior distribution and should be used in Bayesian analyses of phylogeny. In the latter case, branch-length priors can have strong effects on resulting inferences and should be carefully chosen to reflect biological expectations. We provide a formula to calculate an exponential rate parameter for the branch-length prior that should eliminate inference of biased branch lengths in many cases. In any phylogenetic analysis, the biological plausibility of branch-length output must be carefully considered.

Newly discovered sister lineage sheds light on early ant evolution

Ants are the worlds most conspicuous and important eusocial insects and their diversity, abundance, and extreme behavioral specializations make them a model system for several disciplines within the biological sciences. Here, we report the discovery of a new ant that appears to represent the sister lineage to all extant ants (Hymenoptera: Formicidae). The phylogenetic position of this cryptic predator from the soils of the Amazon rainforest was inferred from several nuclear genes, sequenced from a single leg. Martialis heureka (gen. et sp. nov.) also constitutes the sole representative of a new, morphologically distinct subfamily of ants, the Martialinae (subfam. nov.). Our analyses have reduced the likelihood of long-branch attraction artifacts that have troubled previous phylogenetic studies of early-diverging ants and therefore solidify the emerging view that the most basal extant ant lineages are cryptic, hypogaeic foragers. On the basis of morphological and phylogenetic evidence we suggest that these specialized subterranean predators are the sole surviving representatives of a highly divergent lineage that arose near the dawn of ant diversification and have persisted in ecologically stable environments like tropical soils over great spans of time.

Source identification in two criminal cases using phylogenetic analysis of HIV-1 DNA sequences

Phylogenetic analysis has been widely used to test the a priori hypothesis of epidemiological clustering in suspected transmission chains of HIV-1. Among studies showing strong support for relatedness between HIV samples obtained from infected individuals, evidence for the direction of transmission between epidemiologically related pairs has been lacking. During transmission of HIV, a genetic bottleneck occurs, resulting in the paraphyly of source viruses with respect to those of the recipient. This paraphyly establishes the direction of transmission, from which the source can then be inferred. Here, we present methods and results from two criminal cases, State of Washington v Anthony Eugene Whitfield, case number 04-1-0617-5 (Superior Court of the State of Washington, Thurston County, 2004) and State of Texas v Philippe Padieu, case numbers 219-82276-07, 219-82277-07, 219-82278-07, 219-82279-07, 219-82280-07, and 219-82705-07 (219th Judicial District Court, Collin County, TX, 2009), which provided evidence that direction can be established from blinded case samples. The observed paraphyly from each case study led to the identification of an inferred source (i.e., index case), whose identity was revealed at trial to be that of the defendant.

Poor Fit to the Multispecies Coalescent is Widely Detectable in Empirical Data

Model checking is a critical part of Bayesian data analysis, yet it remains largely unused in systematic studies. Phylogeny estimation has recently moved into an era of increasingly complex models that simultaneously account for multiple evolutionary processes, the statistical fit of these models to the data has rarely been tested. Here we develop a posterior predictive simulation-based model check for a commonly used multispecies coalescent model, implemented in *BEAST, and apply it to 25 published data sets. We show that poor model fit is detectable in the majority of data sets; that this poor fit can mislead phylogenetic estimation; and that in some cases it stems from processes of inherent interest to systematists. We suggest that as systematists scale up to phylogenomic data sets, which will be subject to a heterogeneous array of evolutionary processes, critically evaluating the fit of models to data is an analytical step that can no longer be ignored.

Human adipose tissue as a reservoir for memory CD4+ T cells and HIV

Objective:The objective of this study is to determine whether adipose tissue functions as a reservoir for HIV-1. Design:We examined memory CD4+ T cells and HIV DNA in adipose tissue–stromal vascular fraction (AT-SVF) of five patients [four antiretroviral therapy (ART)-treated and one untreated]. To determine whether adipocytes stimulate CD4+ T cells and regulate HIV production, primary human adipose cells were cocultured with HIV-infected CD4+ T cells. Methods:AT-SVF T cells were studied by flow cytometry, and AT-SVF HIV DNA (Gag and Env) was examined by nested PCR and sequence analyses. CD4+ T-cell activation and HIV production were measured by flow cytometry and ELISA. Results:AT-SVF CD3+ T cells were activated (>60% CD69+) memory CD4+ and CD8+ T cells in uninfected and HIV-infected persons, but the AT-SVF CD4+/CD8+ ratio was lower in HIV patients. HIV DNA (Gag and Env) was detected in AT-SVF of all five patients examined by nested PCR, comparably to other tissues [peripheral blood mononuclear cell (PBMC), lymph node or thymus]. In coculture experiments, adipocytes increased CD4+ T-cell activation and HIV production approximately two to three-fold in synergy with gamma-chain cytokines interleukin (IL)-2, IL7 or IL15. These effects were mitigated by neutralizing antibodies against IL6 and integrin-&agr;1&bgr;1. Adipocytes also enhanced T-cell viability. Conclusion:Adipose tissues of ART-treated patients harbour activated memory CD4+ T cells and HIV DNA. Adipocytes promote CD4+ T-cell activation and HIV production in concert with intrinsic adipose factors. Adipose tissue may be an important reservoir for HIV.

A phylogenomic approach to vertebrate phylogeny supports a turtle-archosaur affinity and a possible paraphyletic lissamphibia.

In resolving the vertebrate tree of life, two fundamental questions remain: 1) what is the phylogenetic position of turtles within amniotes, and 2) what are the relationships between the three major lissamphibian (extant amphibian) groups? These relationships have historically been difficult to resolve, with five different hypotheses proposed for turtle placement, and four proposed branching patterns within Lissamphibia. We compiled a large cDNA/EST dataset for vertebrates (75 genes for 129 taxa) to address these outstanding questions. Gene-specific phylogenetic analyses revealed a great deal of variation in preferred topology, resulting in topologically ambiguous conclusions from the combined dataset. Due to consistent preferences for the same divergent topologies across genes, we suspected systematic phylogenetic error as a cause of some variation. Accordingly, we developed and tested a novel statistical method that identifies sites that have a high probability of containing biased signal for a specific phylogenetic relationship. After removing putatively biased sites, support emerged for a sister relationship between turtles and either crocodilians or archosaurs, as well as for a caecilian-salamander sister relationship within Lissamphibia, with Lissamphibia potentially paraphyletic.

Detection of Implausible Phylogenetic Inferences Using Posterior Predictive Assessment of Model Fit

Systematic phylogenetic error caused by the simplifying assumptions made in models of molecular evolution may be impossible to avoid entirely when attempting to model evolution across massive, diverse data sets. However, not all deficiencies of inference models result in unreliable phylogenetic estimates. The field of phylogenetics lacks a direct method to identify cases where model specification adversely affects inferences. Posterior predictive simulation is a flexible and intuitive approach for assessing goodness-of-fit of the assumed model and priors in a Bayesian phylogenetic analysis. Here, I propose new test statistics for use in posterior predictive assessment of model fit. These test statistics compare phylogenetic inferences from posterior predictive data sets to inferences from the original data. A simulation study demonstrates the utility of these new statistics. The new tests reject the plausibility of inferred tree lengths or topologies more often when data/model combinations produce biased inferences. I also apply this approach to exemplar empirical data sets, highlighting the value of the novel assessments.

Can We Identify Genes with Increased Phylogenetic Reliability

Topological heterogeneity among gene trees is widely observed in phylogenomic analyses and some of this variation is likely caused by systematic error in gene tree estimation. Systematic error can be mitigated by improving models of sequence evolution to account for all evolutionary processes relevant to each gene or identifying those genes whose evolution best conforms to existing models. However, the best method for identifying such genes is not well established. Here, we ask if filtering genes according to their clock-likeness or posterior predictive effect size (PPES, an inference-based measure of model violation) improves phylogenetic reliability and congruence. We compared these approaches to each other, and to the common practice of filtering based on rate of evolution, using two different metrics. First, we compared gene-tree topologies to accepted reference topologies. Second, we examined topological similarity among gene trees in filtered sets. Our results suggest that filtering genes based on clock-likeness and PPES can yield a collection of genes with more reliable phylogenetic signal. For the two exemplar data sets we explored, from yeast and amniotes, clock-likeness and PPES outperformed rate-based filtering in both congruence and reliability.