RESUMO
The fossilized birth-death (FBD) model is a naturally appealing way of directly incorporating fossil information when estimating diversification rates. However, an important yet often overlooked property of the original FBD derivation is that it distinguishes between two types of sampled lineages. Here, we first discuss and demonstrate the impact of severely undersampling, and even not including fossils that represent samples of lineages that also had sampled descendants. We then explore the benefits of including fossils, generally, by implementing and then testing two types of FBD models, including one that converts a fossil set into stratigraphic ranges, in more complex likelihood-based models that assume multiple rate classes across the tree. Under various simulation scenarios, including a scenario that exists far outside the set of models we evaluated, including fossils rarely outperform analyses that exclude them altogether. At best, the inclusion of fossils improves precision but does not influence bias. Similarly, we found that converting the fossil set to stratigraphic ranges, which is one way to remedy the effects of undercounting the number of k-type fossils, results in turnover rates and extinction fraction estimates that are generally underestimated. Although fossils remain essential for understanding diversification through time, in the specific case of understanding diversification given an existing, largely modern tree, they are not especially beneficial. [Fossilized birth-death; fossils; MiSSE; state speciation extinction; stratigraphic ranges; turnover rate.].
Assuntos
Fósseis , Especiação Genética , Filogenia , Funções Verossimilhança , TempoRESUMO
Applications of molecular phylogenetic approaches have uncovered evidence of hybridization across numerous clades of life, yet the environmental factors responsible for driving opportunities for hybridization remain obscure. Verbal models implicating geographic range shifts that brought species together during the Pleistocene have often been invoked, but quantitative tests using paleoclimatic data are needed to validate these models. Here, we produce a phylogeny for Heuchereae, a clade of 15 genera and 83 species in Saxifragaceae, with complete sampling of recognized species, using 277 nuclear loci and nearly complete chloroplast genomes. We then employ an improved framework with a coalescent simulation approach to test and confirm previous hybridization hypotheses and identify one new intergeneric hybridization event. Focusing on the North American distribution of Heuchereae, we introduce and implement a newly developed approach to reconstruct potential past distributions for ancestral lineages across all species in the clade and across a paleoclimatic record extending from the late Pliocene. Time calibration based on both nuclear and chloroplast trees recovers a mid- to late-Pleistocene date for most inferred hybridization events, a timeframe concomitant with repeated geographic range restriction into overlapping refugia. Our results indicate an important role for past episodes of climate change, and the contrasting responses of species with differing ecological strategies, in generating novel patterns of range contact among plant communities and therefore new opportunities for hybridization. The new ancestral niche method flexibly models the shape of niche while incorporating diverse sources of uncertainty and will be an important addition to the current comparative methods toolkit. [Ancestral niche reconstruction; hybridization; paleoclimate; pleistocene.].
Assuntos
Hibridização Genética , Filogenia , Filogeografia , Teorema de BayesRESUMO
Ultraconserved elements (UCEs) are stretches of hundreds of nucleotides with highly conserved cores flanked by variable regions. Although the selective forces responsible for the preservation of UCEs are unknown, they are nonetheless believed to contain phylogenetically meaningful information from deep to shallow divergence events. Phylogenetic applications of UCEs assume the same degree of rate heterogeneity applies across the entire locus, including variable flanking regions. We present a Wright-Fisher model of selection on nucleotides (SelON) which includes the effects of mutation, drift, and spatially varying, stabilizing selection for an optimal nucleotide sequence. The SelON model assumes the strength of stabilizing selection follows a position-dependent Gaussian function whose exact shape can vary between UCEs. We evaluate SelON by comparing its performance to a simpler and spatially invariant GTR+Γ model using an empirical data set of 400 vertebrate UCEs used to determine the phylogenetic position of turtles. We observe much improvement in model fit of SelON over the GTR+Γ model, and support for turtles as sister to lepidosaurs. Overall, the UCE-specific parameters SelON estimates provide a compact way of quantifying the strength and variation in selection within and across UCEs. SelON can also be extended to include more realistic mapping functions between sequence and stabilizing selection as well as allow for greater levels of rate heterogeneity. By more explicitly modeling the nature of selection on UCEs, SelON and similar approaches can be used to better understand the biological mechanisms responsible for their preservation across highly divergent taxa and long evolutionary time scales.
Assuntos
Modelos Genéticos , Seleção Genética , Sequência de Bases , Sequência Conservada , FilogeniaRESUMO
In 1974, G. Ledyard Stebbins provided a metaphor illustrating how spatial gradients of biodiversity observed today are by-products of the way environment-population interactions drive species diversification through time. We revisit the narrative behind Stebbins's "cradles" and "museums" of biodiversity to debate two points. First, the usual high-speciation versus low-extinction and tropical versus temperate dichotomies are oversimplifications of the original metaphor and may obscure how gradients of diversity are formed. Second, the way in which we use modern gradients of biodiversity to interpret the potential historical processes that generated them are often still biased by the reasons that motivated Stebbins to propose his original metaphor. Specifically, the field has not yet abandoned the idea that species-rich areas and "basal lineages" indicate centers of origin, nor has it fully appreciated the role of traits as regulators of environment-population dynamics. We acknowledge that the terms "cradles" and "museums" are popular in the literature and that terminologies can evolve with the requirements of the field. However, we also argue that the concepts of cradles and museums have outlived their utility in studies of biogeography and macroevolution and should be replaced by discussions of actual processes at play.
Assuntos
Biodiversidade , Museus , Especiação Genética , Filogenia , Dinâmica PopulacionalRESUMO
BACKGROUND: For decades, codon usage has been used as a measure of adaptation for translational efficiency and translation accuracy of a gene's coding sequence. These patterns of codon usage reflect both the selective and mutational environment in which the coding sequences evolved. Over this same period, gene transfer between lineages has become widely recognized as an important biological phenomenon. Nevertheless, most studies of codon usage implicitly assume that all genes within a genome evolved under the same selective and mutational environment, an assumption violated when introgression occurs. In order to better understand the effects of introgression on codon usage patterns and vice versa, we examine the patterns of codon usage in Lachancea kluyveri, a yeast which has experienced a large introgression. We quantify the effects of mutation bias and selection for translation efficiency on the codon usage pattern of the endogenous and introgressed exogenous genes using a Bayesian mixture model, ROC SEMPPR, which is built on mechanistic assumptions about protein synthesis and grounded in population genetics. RESULTS: We find substantial differences in codon usage between the endogenous and exogenous genes, and show that these differences can be largely attributed to differences in mutation bias favoring A/T ending codons in the endogenous genes while favoring C/G ending codons in the exogenous genes. Recognizing the two different signatures of mutation bias and selection improves our ability to predict protein synthesis rate by 42% and allowed us to accurately assess the decaying signal of endogenous codon mutation and preferences. In addition, using our estimates of mutation bias and selection, we identify Eremothecium gossypii as the closest relative to the exogenous genes, providing an alternative hypothesis about the origin of the exogenous genes, estimate that the introgression occurred â¼6×108 generation ago, and estimate its historic and current selection against mismatched codon usage. CONCLUSIONS: Our work illustrates how mechanistic, population genetic models like ROC SEMPPR can separate the effects of mutation and selection on codon usage and provide quantitative estimates from sequence data.
Assuntos
Uso do Códon , Genética Populacional , Modelos Genéticos , Saccharomycetales/genética , Seleção Genética , Teorema de Bayes , MutaçãoRESUMO
BACKGROUND: Researchers often measure changes in gene expression across conditions to better understand the shared functional roles and regulatory mechanisms of different genes. Analogous to this is comparing gene expression across species, which can improve our understanding of the evolutionary processes shaping the evolution of both individual genes and functional pathways. One area of interest is determining genes showing signals of coevolution, which can also indicate potential functional similarity, analogous to co-expression analysis often performed across conditions for a single species. However, as with any trait, comparing gene expression across species can be confounded by the non-independence of species due to shared ancestry, making standard hypothesis testing inappropriate. RESULTS: We compared RNA-Seq data across 18 fungal species using a multivariate Brownian Motion phylogenetic comparative method (PCM), which allowed us to quantify coevolution between protein pairs while directly accounting for the shared ancestry of the species. Our work indicates proteins which physically-interact show stronger signals of coevolution than randomly-generated pairs. Interactions with stronger empirical and computational evidence also showing stronger signals of coevolution. We examined the effects of number of protein interactions and gene expression levels on coevolution, finding both factors are overall poor predictors of the strength of coevolution between a protein pair. Simulations further demonstrate the potential issues of analyzing gene expression coevolution without accounting for shared ancestry in a standard hypothesis testing framework. Furthermore, our simulations indicate the use of a randomly-generated null distribution as a means of determining statistical significance for detecting coevolving genes with phylogenetically-uncorrected correlations, as has previously been done, is less accurate than PCMs, although is a significant improvement over standard hypothesis testing. These methods are further improved by using a phylogenetically-corrected correlation metric. CONCLUSIONS: Our work highlights potential benefits of using PCMs to detect gene expression coevolution from high-throughput omics scale data. This framework can be built upon to investigate other evolutionary hypotheses, such as changes in transcription regulatory mechanisms across species.
Assuntos
Evolução Molecular , Proteínas Fúngicas/genética , Fungos/genética , Expressão Gênica , Proteínas Fúngicas/metabolismo , Fungos/classificação , Fungos/metabolismo , Modelos Genéticos , Fenótipo , Filogenia , Ligação ProteicaRESUMO
We present a new phylogenetic approach, selection on amino acids and codons (SelAC), whose substitution rates are based on a nested model linking protein expression to population genetics. Unlike simpler codon models that assume a single substitution matrix for all sites, our model more realistically represents the evolution of protein-coding DNA under the assumption of consistent, stabilizing selection using a cost-benefit approach. This cost-benefit approach allows us to generate a set of 20 optimal amino acid-specific matrix families using just a handful of parameters and naturally links the strength of stabilizing selection to protein synthesis levels, which we can estimate. Using a yeast data set of 100 orthologs for 6 taxa, we find SelAC fits the data much better than popular models by 104-105 Akike information criterion units adjusted for small sample bias. Our results also indicated that nested, mechanistic models better predict observed data patterns highlighting the improvement in biological realism in amino acid sequence evolution that our model provides. Additional parameters estimated by SelAC indicate that a large amount of nonphylogenetic, but biologically meaningful, information can be inferred from existing data. For example, SelAC prediction of gene-specific protein synthesis rates correlates well with both empirical (r=0.33-0.48) and other theoretical predictions (r=0.45-0.64) for multiple yeast species. SelAC also provides estimates of the optimal amino acid at each site. Finally, because SelAC is a nested approach based on clearly stated biological assumptions, future modifications, such as including shifts in the optimal amino acid sequence within or across lineages, are possible.
Assuntos
Substituição de Aminoácidos , Técnicas Genéticas , Modelos Genéticos , Filogenia , Seleção Genética , Genética Populacional/métodosRESUMO
Leaf reflectance spectra have been increasingly used to assess plant diversity. However, we do not yet understand how spectra vary across the tree of life or how the evolution of leaf traits affects the differentiation of spectra among species and lineages. Here we describe a framework that integrates spectra with phylogenies and apply it to a global dataset of over 16 000 leaf-level spectra (400-2400 nm) for 544 seed plant species. We test for phylogenetic signal in spectra, evaluate their ability to classify lineages, and characterize their evolutionary dynamics. We show that phylogenetic signal is present in leaf spectra but that the spectral regions most strongly associated with the phylogeny vary among lineages. Despite among-lineage heterogeneity, broad plant groups, orders, and families can be identified from reflectance spectra. Evolutionary models also reveal that different spectral regions evolve at different rates and under different constraint levels, mirroring the evolution of their underlying traits. Leaf spectra capture the phylogenetic history of seed plants and the evolutionary dynamics of leaf chemistry and structure. Consequently, spectra have the potential to provide breakthrough assessments of leaf evolution and plant phylogenetic diversity at global scales.
Assuntos
Folhas de Planta , Sementes , Filogenia , PlantasRESUMO
Early flowering plants are thought to have been woody species restricted to warm habitats. This lineage has since radiated into almost every climate, with manifold growth forms. As angiosperms spread and climate changed, they evolved mechanisms to cope with episodic freezing. To explore the evolution of traits underpinning the ability to persist in freezing conditions, we assembled a large species-level database of growth habit (woody or herbaceous; 49,064 species), as well as leaf phenology (evergreen or deciduous), diameter of hydraulic conduits (that is, xylem vessels and tracheids) and climate occupancies (exposure to freezing). To model the evolution of species' traits and climate occupancies, we combined these data with an unparalleled dated molecular phylogeny (32,223 species) for land plants. Here we show that woody clades successfully moved into freezing-prone environments by either possessing transport networks of small safe conduits and/or shutting down hydraulic function by dropping leaves during freezing. Herbaceous species largely avoided freezing periods by senescing cheaply constructed aboveground tissue. Growth habit has long been considered labile, but we find that growth habit was less labile than climate occupancy. Additionally, freezing environments were largely filled by lineages that had already become herbs or, when remaining woody, already had small conduits (that is, the trait evolved before the climate occupancy). By contrast, most deciduous woody lineages had an evolutionary shift to seasonally shedding their leaves only after exposure to freezing (that is, the climate occupancy evolved before the trait). For angiosperms to inhabit novel cold environments they had to gain new structural and functional trait solutions; our results suggest that many of these solutions were probably acquired before their foray into the cold.
Assuntos
Evolução Biológica , Clima Frio , Ecossistema , Congelamento , Magnoliopsida/anatomia & histologia , Magnoliopsida/fisiologia , Xilema/anatomia & histologia , Funções Verossimilhança , Filogeografia , Folhas de Planta/anatomia & histologia , Folhas de Planta/fisiologia , Sementes/fisiologia , Fatores de Tempo , Madeira/anatomia & histologia , Madeira/fisiologia , Xilema/fisiologiaRESUMO
The demographic history of most species is complex, with multiple evolutionary processes combining to shape the observed patterns of genetic diversity. To infer this history, the discipline of phylogeography has (to date) used models that simplify the historical demography of the focal organism, for example by assuming or ignoring ongoing gene flow between populations or by requiring a priori specification of divergence history. Since no single model incorporates every possible evolutionary process, researchers rely on intuition to choose the models that they use to analyze their data. Here, we describe an approximate likelihood approach that reduces this reliance on intuition. PHRAPL allows users to calculate the probability of a large number of complex demographic histories given a set of gene trees, enabling them to identify the most likely underlying model and estimate parameters for a given system. Available model parameters include coalescence time among populations or species, gene flow, and population size. We describe the method and test its performance in model selection and parameter estimation using simulated data. We also compare model probabilities estimated using our approximate likelihood method to those obtained using standard analytical likelihood. The method performs well under a wide range of scenarios, although this is sometimes contingent on sampling many loci. In most scenarios, as long as there are enough loci and if divergence among populations is sufficiently deep, PHRAPL can return the true model in nearly all simulated replicates. Parameter estimates from the method are also generally accurate in most cases. PHRAPL is a valuable new method for phylogeographic model selection and will be particularly useful as a tool to more extensively explore demographic model space than is typically done or to estimate parameters for complex models that are not readily implemented using current methods. Estimating relevant parameters using the most appropriate demographic model can help to sharpen our understanding of the evolutionary processes giving rise to phylogeographic patterns. [AIC; grid search; isolation-with-migration; migration rate; multispecies coalescent; parameter optimization; population genetics; tree topologies.].
Assuntos
Modelos Biológicos , Filogeografia/métodos , Evolução Biológica , Simulação por Computador , Variação Genética , Funções Verossimilhança , FilogeniaRESUMO
Species are commonly thought to be evolutionarily independent in a way that populations within a species are not. In recent years, studies that seek to identify evolutionarily independent lineages (i.e., to delimit species) using genetic data have typically adopted multispecies coalescent approaches that assume that evolutionary independence is formed by the differential sorting of ancestral alleles due to genetic drift. However, gene flow appears to be common among populations and nascent species, and while this process may inhibit lineage divergence (and thus independence), it is usually not explicitly considered when delimiting species. In this article, we apply Phylogeographic Inference using Approximate Likelihoods (PHRAPL), a recently described method for phylogeographic model selection, to species delimitation. We describe an approach to delimiting species using PHRAPL that attempts to account for both genetic drift and gene flow, and we compare the method's performance to that of a popular delimitation approach (BPP) using both simulated and empirical datasets. PHRAPL generally infers the correct demographic-delimitation model when the generating model includes gene flow between taxa, given a sufficient amount of data. When the generating model includes only isolation in the recent past, PHRAPL will in some cases fail to differentiate between gene flow and divergence, leading to model misspecification. Nevertheless, the explicit consideration of gene flow by PHRAPL is an important complement to existing delimitation approaches, particularly in systems where gene flow is likely important. [approximate likelihoods; coalescent simulations; genealogical divergence index; Homo sapiens; isolation-with-migration; multispecies coalescent; Sarracenia; Scincella.].
Assuntos
Classificação/métodos , Fluxo Gênico , Especiação Genética , Modelos Biológicos , Filogenia , Teorema de Bayes , Simulação por Computador , Filogeografia , Especificidade da EspécieRESUMO
Growing evidence supports the idea that species can diverge in the presence of gene flow. However, most methods of phylogeny estimation do not consider this process, despite the fact that ignoring gene flow is known to bias phylogenetic inference. Furthermore, studies that do consider divergence-with-gene-flow typically do so by estimating rates of gene flow using a isolation-with-migration model (IM), rather than evaluating scenarios of gene flow (such as divergence-with-gene flow or secondary contact) that represent very different types of diversification. In this investigation, we aim to infer the recent phylogenetic history of a clade of western long-eared bats while evaluating a number of different models that parameterize gene flow in a variety of ways. We utilize PHRAPL, a new tool for phylogeographic model selection, to compare the fit of a broad set of demographic models that include divergence, migration, or both among Myotis evotis, $M$. thysanodes and M. keenii. A genomic data set consisting of 808 loci of ultraconserved elements was used to explore such models in three steps using an incremental design where each successive set was informed by, and thus more focused than, the previous set of models. Specifically, the three steps were to (i) assess whether gene flow should be modeled and identify the best topologies, (ii) infer directionality of migration using the best topologies, and (iii) estimate the timing of gene flow. The best model (AIC model weight ${\sim}0.98$) included two divergence events (($M$. evotis, $M$. thysanodes), M. keenii) accompanied by gene flow at the initial stages of divergence. These results provide a striking example of speciation-with-gene-flow in an evolutionary lineage. [Myotis bats; PHRAPL; P2C2M; phylogeographic model selection; speciation with gene flow.].
Assuntos
Quirópteros/classificação , Quirópteros/genética , Fluxo Gênico , Especiação Genética , Filogenia , Animais , Evolução Biológica , Modelos Biológicos , Software , Estados UnidosRESUMO
PREMISE OF THE STUDY: The study of very large and very old clades holds the promise of greater insights into evolution across the tree of life. However, there has been a fair amount of criticism regarding the interpretations and quality of studies to date, with some suggesting that detailed studies carried out on smaller, tractable scales should be preferred over the increasingly grand syntheses of these data. METHODS: We provided in detail our trials and tribulations of compiling a large, sparsely sampled matrix from GenBank data and inferring a well-supported, time-calibrated phylogeny of Campanulidae. We also used a simulation approach to assess tree quality and to study the value of using very large, comprehensive phylogenies in a comparative context. KEY RESULTS: A robust and well-supported phylogeny can be produced as long as automated procedures are supplemented with some human intervention. In the case of campanulids, the overall topology may be driven not only by particular genes, but also particular sequences for a gene. We also determined that estimates of divergence times should be fairly robust to issues related to clade-specific heterogeneity. Finally, we demonstrated how relying on results from smaller, younger clades are prone to produce biased interpretations of tropical to temperate evolution across campanulids as a whole. CONCLUSIONS: While we were both surprised and encouraged by the robust and fairly well-resolved, comprehensive phylogeny of campanulids, challenges still remain. Nevertheless, large phylogenies are inherently valuable in a comparative context if only to attenuate the issue of ascertainment bias.
Assuntos
Sequência de Bases , Evolução Biológica , DNA de Plantas/análise , Genes de Plantas , Magnoliopsida/genética , Filogenia , Evolução Molecular , Análise de Sequência de DNARESUMO
Phylogeography seeks to discover the evolutionary processes that have given rise to organismal and genetic diversity. This requires explicit hypotheses (i.e., models) to be evaluated with genetic data in order to identify those hypotheses that best explain the data. In recent years, advancements in the model-based tools used to estimate phylogeographic parameters of interest such as gene flow, divergence time, and relationships among groups have been made. However, given the complexity of these models, available methods can typically only compare a handful of possible hypotheses, requiring researchers to specify in advance the small set of models to consider. Without formal quantitative approaches to model selection, researchers must rely on their intuition to formulate the model space to be explored. We explore the adequacy of intuitive choices made by researchers during the process of data analysis by reanalyzing 20 empirical phylogeographic datasets using PHRAPL, an objective tool for phylogeographic model selection. We show that the best models for most datasets include both gene flow and population divergence parameters, and that species tree methods (which do not consider gene flow) tend to be overly simplistic for many phylogeographic systems. Objective approaches to phylogeographic model selection offer an important complement to researcher intuition.
Assuntos
Modelos Genéticos , Filogeografia , Animais , Variação Genética , Filogenia , ProbabilidadeRESUMO
The distribution of diversity can vary considerably from clade to clade. Attempts to understand these patterns often employ state-dependent speciation and extinction models to determine whether the evolution of a particular novel trait has increased speciation rates and/or decreased extinction rates. It is still unclear, however, whether these models are uncovering important drivers of diversification, or whether they are simply pointing to more complex patterns involving many unmeasured and co-distributed factors. Here we describe an extension to the popular state-dependent speciation and extinction models that specifically accounts for the presence of unmeasured factors that could impact diversification rates estimated for the states of any observed trait, addressing at least one major criticism of BiSSE (Binary State Speciation and Extinction) methods. Specifically, our model, which we refer to as HiSSE (Hidden State Speciation and Extinction), assumes that related to each observed state in the model are "hidden" states that exhibit potentially distinct diversification dynamics and transition rates than the observed states in isolation. We also demonstrate how our model can be used as character-independent diversification models that allow for a complex diversification process that is independent of the evolution of a character. Under rigorous simulation tests and when applied to empirical data, we find that HiSSE performs reasonably well, and can at least detect net diversification rate differences between observed and hidden states and detect when diversification rate differences do not correlate with the observed states. We discuss the remaining issues with state-dependent speciation and extinction models in general, and the important ways in which HiSSE provides a more nuanced understanding of trait-dependent diversification.
Assuntos
Classificação/métodos , Extinção Biológica , Especiação Genética , Modelos Biológicos , Biodiversidade , Fenótipo , FilogeniaRESUMO
Why are some traits and trait combinations exceptionally common across the tree of life, whereas others are vanishingly rare? The distribution of trait diversity across a clade at any time depends on the ancestral state of the clade, the rate at which new phenotypes evolve, the differences in speciation and extinction rates across lineages, and whether an equilibrium has been reached. Here we examine the role of transition rates, differential diversification (speciation minus extinction) and non-equilibrium dynamics on the evolutionary history of angiosperms, a clade well known for the abundance of some trait combinations and the rarity of others. Our analysis reveals that three character states (corolla present, bilateral symmetry, reduced stamen number) act synergistically as a key innovation, doubling diversification rates for lineages in which this combination occurs. However, this combination is currently less common than predicted at equilibrium because the individual characters evolve infrequently. Simulations suggest that angiosperms will remain far from the equilibrium frequencies of character states well into the future. Such non-equilibrium dynamics may be common when major innovations evolve rarely, allowing lineages with ancestral forms to persist, and even outnumber those with diversification-enhancing states, for tens of millions of years.
Assuntos
Flores/anatomia & histologia , Flores/fisiologia , Magnoliopsida/fisiologia , Biodiversidade , Evolução Biológica , Modelos Biológicos , Filogenia , Característica Quantitativa Herdável , Processos EstocásticosRESUMO
Dating analyses based on molecular data imply that crown angiosperms existed in the Triassic, long before their undisputed appearance in the fossil record in the Early Cretaceous. Following a re-analysis of the age of angiosperms using updated sequences and fossil calibrations, we use a series of simulations to explore the possibility that the older age estimates are a consequence of (i) major shifts in the rate of sequence evolution near the base of the angiosperms and/or (ii) the representative taxon sampling strategy employed in such studies. We show that both of these factors do tend to yield substantially older age estimates. These analyses do not prove that younger age estimates based on the fossil record are correct, but they do suggest caution in accepting the older age estimates obtained using current relaxed-clock methods. Although we have focused here on the angiosperms, we suspect that these results will shed light on dating discrepancies in other major clades.
Assuntos
Evolução Molecular , Magnoliopsida/classificação , Magnoliopsida/genética , Filogenia , Simulação por Computador , Fósseis , Modelos Genéticos , TempoRESUMO
The popular likelihood-based model selection criterion, Akaike's Information Criterion (AIC), is a breakthrough mathematical result derived from information theory. AIC is an approximation to Kullback-Leibler (KL) divergence with the derivation relying on the assumption that the likelihood function has finite second derivatives. However, for phylogenetic estimation, given that tree space is discrete with respect to tree topology, the assumption of a continuous likelihood function with finite second derivatives is violated. In this paper, we investigate the relationship between the expected log likelihood of a candidate model, and the expected KL divergence in the context of phylogenetic tree estimation. We find that given the tree topology, AIC is an unbiased estimator of the expected KL divergence. However, when the tree topology is unknown, AIC tends to underestimate the expected KL divergence for phylogenetic models. Simulation results suggest that the degree of underestimation varies across phylogenetic models so that even for large sample sizes, the bias of AIC can result in selecting a wrong model. As the choice of phylogenetic models is essential for statistical phylogenetic inference, it is important to improve the accuracy of model selection criteria in the context of phylogenetics.