Fact-checked by Grok 2 weeks ago

Phylogenetic comparative methods

Phylogenetic comparative methods (PCMs) are statistical approaches in that utilize phylogenetic trees—diagrams representing the evolutionary relationships among —to analyze patterns of and diversification while correcting for the non-independence of arising from shared ancestry. Introduced to overcome limitations in traditional analyses that ignored phylogenetic , PCMs enable researchers to test hypotheses about evolutionary processes, such as , , and the tempo and mode of change. The foundational work on PCMs dates to 1985, when Joseph Felsenstein proposed phylogenetically contrasts (PIC), a method that transforms trait data into a set of statistically contrasts along the branches of a phylogeny, allowing valid and analyses. This addressed the problem in cross-species comparisons, where closely related species share similar traits due to rather than evolution. Since then, the field has expanded rapidly, with publications on PCMs increasing dramatically from the onward, reflecting advances in and the availability of large-scale molecular data. Key extensions include phylogenetic (PGLS), which incorporates phylogenetic covariance into linear models, and stochastic models of trait evolution such as (random drift) and the Ornstein-Uhlenbeck process (which includes toward an optimum). PCMs are applied across diverse domains to investigate how traits evolve in response to ecological, genetic, or environmental factors, including studies of morphological adaptations in , physiological traits in , and even cultural evolution in societies. For instance, they have been used to examine the evolution of nitrogen-fixing symbioses in angiosperms and political in Austronesian languages. Developments since 2020 have further integrated PCMs with meta-analyses for decomposing variation in evolutionary studies, phylogenetic genotype-to-phenotype mapping for predicting traits from genomes, and applications in to leverage comparative for health insights. Despite their power, PCMs require careful consideration of assumptions like accurate phylogenies and appropriate evolutionary models, as violations can lead to biased inferences; as of 2025, ongoing efforts continue to improve accessibility through updated R packages such as phytools (version 2.0, 2024) and OUwie (2025), while fostering better communication between method developers and users.

Introduction

Definition and scope

Phylogenetic comparative methods (PCMs) are a class of statistical techniques designed to analyze comparative data across species while explicitly incorporating their phylogenetic relationships to account for non-independence arising from shared evolutionary ancestry. These methods address the fundamental problem that species are not independent data points in evolutionary analyses, as closely related species tend to resemble each other more due to rather than independent . By integrating phylogenetic information, PCMs enable researchers to make inferences about evolutionary processes, such as trait evolution and adaptation, that would otherwise be confounded by ignoring historical contingencies. The scope of PCMs extends broadly across , encompassing analyses of trait correlations, rates of evolutionary change, patterns of species diversification, and tests of to environmental pressures. They are particularly valuable for investigating macroevolutionary hypotheses, such as whether certain traits drive adaptive radiations or influence and dynamics over . For instance, PCMs facilitate the study of how continuous traits like body size evolve and correlate with other variables, such as metabolic rate, across clades like mammals, by modeling changes along phylogenetic branches to distinguish phylogenetic signal from ecological drivers. A key prerequisite for applying PCMs is the availability of a phylogeny, typically represented as a with (branching order) and branch lengths (indicating evolutionary time or amounts), which serves as the foundational structure for incorporating shared ancestry into statistical models. This input allows PCMs to transform correlated species-level into phylogenetically informed contrasts or simulations, thereby providing a robust framework for testing in .

Historical background

The roots of phylogenetic comparative methods (PCMs) trace back to 19th-century , where pioneered systematic comparisons of organismal structures to infer functional relationships, laying groundwork for analyzing evolutionary patterns across species despite his rejection of . In the early , advanced quantitative approaches to interspecific variation; Karl Pearson's development of and techniques, including applications to taxonomic data, provided statistical tools for handling species-level comparisons, though without explicit phylogenetic correction. A pivotal shift occurred in 1985 with Joseph Felsenstein's seminal paper, which highlighted the problem of phylogenetic autocorrelation—non-independence in trait data due to shared ancestry—and proposed phylogenetically independent contrasts as a solution to enable valid statistical inferences in comparative analyses. This work addressed longstanding statistical challenges in comparative biology by integrating phylogenetic trees into regression-like frameworks. The 1990s and 2000s saw rapid expansion, driven by simulation-based validations and new modeling approaches. Researchers such as Ted Garland, Emilia Martins, and Mark Pagel tested method robustness through simulations and introduced innovations like likelihood-based models for correlated evolution and models for traits, broadening PCM applications to diverse evolutionary questions. Post-2010, Bayesian implementations gained prominence, incorporating distributions on evolutionary parameters and phylogenetic to enhance flexibility and accuracy in complex datasets. Recent advances through 2025 have integrated PCMs with phylogenomics to accommodate tree discordance from processes like incomplete sorting, improving estimates on genome-scale trees. Hybrid approaches combining PCMs with have also emerged, enabling scalable analyses of large datasets by leveraging predictive models for pattern detection and parameter optimization.

Key Milestones

Fundamental Concepts

Phylogenetic trees and comparative data

Phylogenetic trees are branching diagrams that depict the evolutionary relationships among a set of taxa, such as or genes, illustrating their descent from common ancestors. The structure of a phylogenetic tree consists of three main components: the , which defines the branching pattern and hierarchical relationships among taxa; branch lengths, which represent the amount of evolutionary change, such as time since divergence or number of genetic substitutions; and polytomies, which are nodes with more than two immediate descendant branches, indicating unresolved or simultaneous divergences due to insufficient data or rapid events. These trees serve as the foundational framework for phylogenetic comparative methods (PCMs), enabling analyses that account for shared evolutionary history among taxa. Phylogenetic trees are estimated using various statistical approaches, including maximum likelihood methods implemented in software like RAxML, which optimize tree topologies and branch lengths based on probabilistic models of evolution, and via programs such as MrBayes, which incorporate prior probabilities to sample from the posterior distribution of trees. In PCMs, the choice between ultrametric trees—where all tips are equidistant from the , assuming a constant rate of ()—and non-ultrametric trees, which allow varying evolutionary rates across branches, is critical; ultrametric trees are often preferred for time-calibrated analyses, while non-ultrametric trees better capture in studies without strict clock assumptions. Comparative data in PCMs typically comprise trait measurements collected from extant taxa at the tips of the , known as tip-only sampling, to infer evolutionary processes without direct observations of ancestral states. These data include continuous traits, such as body mass or , which vary along a quantitative ; discrete traits, like the presence or absence of specific morphological features (e.g., horns in ungulates); and multivariate datasets combining multiple traits to explore covariation. Aligning comparative data with the phylogenetic tree requires ensuring that trait measurements correspond precisely to the taxa represented in the tree, often necessitating taxonomic to match names or identifiers. , common in comparative s due to incomplete sampling, can be addressed through subsetting—removing taxa with incomplete records to retain only fully observed cases—or imputation methods that leverage phylogenetic relationships and available traits to estimate values, thereby preserving size and statistical power. For instance, in analyses of mammalian , from 1,311 extant compiled on a supertree were used to examine allometric scaling while accounting for evolutionary non-independence.

Statistical challenges in comparative analysis

In comparative analyses of species traits, a primary statistical challenge arises from the non-independence of data points, as species share evolutionary history through common ancestry, violating the independence assumption of conventional statistical methods like ordinary least squares (OLS) regression. This shared ancestry induces correlations among trait values, leading to inflated Type I error rates and overestimation of statistical significance in tests of evolutionary hypotheses. For instance, simulations demonstrate that applying OLS to phylogenetically related species can produce p-values that are orders of magnitude smaller than their true values, falsely rejecting null hypotheses of no trait association. Phylogenetic autocorrelation quantifies this non-independence as the tendency for closely related species to exhibit more similar trait values than expected under a random model of evolution. Metrics such as Moran's I adapt spatial autocorrelation measures to phylogenetic trees, revealing patterned similarity driven by descent rather than ecological convergence. Similarly, Blomberg's K assesses the phylogenetic signal by comparing observed trait variance to that expected under a Brownian motion process of evolution, where values greater than 1 indicate stronger conservatism along branches than predicted. This autocorrelation contributes to pseudoreplication, where multiple species within a clade are treated as independent replicates, artificially increasing sample size and degrees of freedom in analyses. The hierarchical structure of phylogenetic data further complicates , with nested within clades that reflect varying depths of evolutionary , necessitating approaches that account for this multilevel to avoid biased parameter estimates. Violations of key assumptions exacerbate these issues, including uneven sampling across the —where some lineages are overrepresented—leading to distorted inferences about evolution; incomplete phylogenies that omit branches or , introducing uncertainty in structures; and heterogeneity in evolution rates across clades, which can mask or amplify spurious correlations if not addressed. Phylogenetic comparative methods (PCMs) mitigate these challenges by either transforming the data to restore approximate among observations or explicitly modeling the phylogenetic to incorporate the structured non-independence into likelihood-based frameworks. For example, in analyses of length and distance, standard correlations without phylogenetic correction overestimate the strength and significance of the relationship due to shared ancestry among migratory lineages, whereas PCMs reveal more accurate adaptive patterns.

Evolutionary Models

Brownian motion model

The (BM) model is a drift process that serves as the foundational model for continuous in phylogenetic comparative methods, depicting changes as a where increments are normally distributed with mean zero and variance proportional to elapsed time. Introduced by Cavalli-Sforza and Edwards (1967) in the context of gene frequency , the model assumes constant evolutionary rates without directional or , resulting in diffusive changes that accumulate variance linearly over time. Key assumptions of the BM model include the absence of , a constant rate of evolutionary change across all branches of the phylogeny, and normally distributed trait increments with unchanging variance. Under these conditions, values at the tips of a follow a multivariate Gaussian distribution, where the vector reflects the ancestral state and the covariance structure is determined by shared evolutionary . Mathematically, for a evolving along a single of length t, the expected variance is given by \text{Var}(X) = \sigma^2 t, where \sigma^2 represents the evolutionary , denoting the variance of change per time. For two i and j, the between their values is \text{Cov}(X_i, X_j) = \sigma^2 t_{\text{MRCA}}, with t_{\text{MRCA}} as the time since their , capturing the shared path length that induces phylogenetic correlation. This formulation yields a phylogenetic \mathbf{V} whose diagonal elements are \sigma^2 times the total lengths to each , and off-diagonal elements are \sigma^2 times the summed lengths from tips to their common ancestors. In phylogenetic comparative methods, the BM model underpins the expected covariances for analyzing trait correlations across species, enabling corrections for non-independence due to shared ancestry. The rate parameter \sigma^2 is typically estimated using maximum likelihood, which maximizes the probability of observing the tip trait data under the multivariate normal distribution defined by the phylogenetic covariance matrix. Despite its simplicity, the BM model has notable limitations, as it predicts unbounded trait variance that increases indefinitely with divergence time, failing to capture scenarios involving stabilizing selection where traits remain constrained around an optimum. For instance, simulations of a trait evolving under BM on a phylogeny often show variance expanding proportionally along longer branches, leading to dispersed tip values that may unrealistically diverge without limits in real biological systems.

Ornstein-Uhlenbeck process

The process serves as an extension of the model in phylogenetic comparative methods by incorporating a mean-reverting force that pulls trait values toward an optimal θ at rate α, while retaining fluctuations driven by evolutionary rate σ²; this formulation captures under toward selective regimes. Unlike pure drift processes, the OU model allows traits to evolve toward adaptive peaks, with the strength of selection determining the speed of reversion and the degree of phylogenetic signal decay over time. The mathematical foundation of the OU process is given by the stochastic differential equation dX(t) = \alpha \left( \theta - X(t) \right) dt + \sigma \, dW(t), where X(t) is the trait value at time t, \theta is the optimal trait value, \alpha > 0 governs the pull toward the optimum (with phylogenetic half-life \ln 2 / \alpha), \sigma > 0 scales the diffusive noise from the Wiener process W(t), and the process assumes stationarity as t \to \infty. The stationary variance around the optimum is \sigma^2 / (2\alpha), reflecting the balance between selection and stochasticity; for finite times along a phylogeny, the variance at a tip is \frac{\sigma^2}{2\alpha} (1 - e^{-2\alpha t}), where t is the branch length from the root. Key parameters include \theta, the primary adaptive optimum (which can shift across branches to model regime changes); \alpha, the intensity of stabilizing selection; and \sigma^2, the baseline rate of evolutionary fluctuation independent of distance from the optimum. Multiple optima can be specified for different phylogenetic regimes, allowing tests for evolutionary shifts in selective pressures. The covariance between trait values at two tips i and j incorporates exponential decay based on their divergence time \tau and \alpha, typically of the form \frac{\sigma^2}{2\alpha} e^{-\alpha (t_i + t_j - 2\tau)} (1 - e^{-2\alpha \tau}), where t_i and t_j are times from the root; this structure induces phylogenetic correlations that weaken with stronger selection or longer divergence times. Parameter estimation for the OU process relies on likelihood-based methods, maximizing the multivariate normal likelihood of observed tip data under the phylogenetic derived from the model. A seminal application involved fitting the model to body size in lizards across Greater Antillean ecomorphs, where identified multiple optima corresponding to habitat-specific regimes (e.g., perch diameter), with \alpha \approx 0.5–1.0 indicating moderate selection pulling toward ecomorph ideals and \sigma^2 reflecting variation. Advances in the framework include Hansen's 1997 introduction of multi-optima models to detect adaptive shifts without a priori specification of regimes, enabling hypothesis testing via comparisons across models with varying numbers of . Post-2010 developments feature the OUwie package for maximum-likelihood multi-regime fitting alongside Bayesian implementations that integrate phylogenetic uncertainty and prior distributions on parameters, such as and RevBayes, facilitating robust for complex multi-trait or regime-switching scenarios.

Main Statistical Approaches

Phylogenetically independent contrasts

Phylogenetically independent contrasts () transform continuous data across into a set of statistically independent differences, thereby eliminating the non-independence introduced by shared phylogenetic history in comparative analyses. Developed by Felsenstein in , this method extracts contrasts at each internal node of the phylogeny, effectively representing evolutionary changes along branches while accounting for the expected structure under a model of . By focusing on these contrasts rather than raw values, enables the use of conventional statistical tests, such as or , to infer evolutionary relationships without confounding effects from ancestry. The core procedure begins with the tips of the and proceeds upward, computing differences for pairs of sister taxa or clades sharing a common . For two such descendants i and j with trait values X_i and X_j, and branch lengths v_i and v_j (representing expected evolutionary divergence from the ), the standardized contrast is calculated as \delta = \frac{X_i - X_j}{\sqrt{v_i + v_j}} This division by the square root of the sum of branch lengths standardizes the contrasts to have unit expected variance, drawing on the covariances implied by evolution where trait variance accrues proportionally to time or divergence. The process generates n-1 contrasts for a with n , with branch lengths for ancestral nodes updated as the weighted sum of descendant branches (v_k = v_i v_j / (v_i + v_j)) to continue the . For bivariate analyses, contrasts from two traits (\delta_X and \delta_Y) are then regressed through the origin (\delta_Y = \beta \delta_X), where the \beta estimates the evolutionary association free of phylogenetic bias. PIC assumes an accurate phylogeny, typically bifurcating with branch lengths calibrated to reflect expected trait variance under , and that the tree need not be strictly ultrametric but should represent additive distances. It further presumes continuous traits with normally distributed changes, as the method relies on least-squares estimation. To assess the validity of branch length standardization and adherence, a diagnostic regresses the absolute values of contrasts against their standard deviations (\sqrt{v_i + v_j}); a significantly different from zero suggests rate heterogeneity or model misspecification, prompting branch length transformations like logarithmic . In practice, has been applied to test correlations while controlling for phylogeny, such as in analyses of versus body size across . Conventional non-phylogenetic on often overestimates the allometric slope due to similarity among relatives, but PIC contrasts yield an unbiased estimate, typically near 0.6, highlighting the independent of encephalization relative to somatic scaling. Key limitations include high sensitivity to inaccuracies in tree topology or branch lengths, which can inflate Type I error rates or distort contrast variances, leading to erroneous conclusions about trait covariation. Additionally, is inappropriate for traits, as it assumes continuous, normally distributed data and cannot handle categorical or variables without violating its statistical foundations.

Phylogenetic generalized least squares

Phylogenetic generalized least squares (PGLS) extends the (GLS) regression framework to incorporate phylogenetic information, modeling the among species traits as arising from shared evolutionary history. This approach accounts for non-independence in data by estimating a phylogenetic derived from an evolutionary model, allowing for flexible analysis under various assumptions about trait evolution. The core formulation of PGLS is a Y = X\beta + \epsilon, where Y is the response variable across , X is the of predictors, \beta is the of coefficients, and \epsilon is the error term with \text{Var}(\epsilon) = \sigma^2 V. Here, V is the phylogenetic scaled by the error variance \sigma^2, with elements V_{ij} reflecting the shared evolutionary history between i and j; for example, under the (BM) model, V_{ij} = t_{\text{MRCA}}, the time to the of i and j. This matrix V can be derived from any specified evolutionary model, enabling PGLS to adapt to different processes of trait change. Parameter estimation in PGLS typically uses maximum likelihood to obtain \hat{\beta} and associated statistics, with the covariance matrix inverted to weight observations by their phylogenetic relatedness. To refine the model, branch lengths in the phylogeny can be transformed using parameters such as Pagel's \lambda, which scales the off-diagonal elements of V (where \lambda = 0 assumes a star phylogeny with no covariances, and \lambda = 1 corresponds to the full model), or Blomberg et al.'s \kappa, which raises branch lengths to a power to alter the tempo of . The ordinary estimator is modified to \hat{\beta} = (X' V^{-1} X)^{-1} X' V^{-1} Y, providing unbiased estimates of slopes and intercepts while accounting for phylogenetic structure. For instance, PGLS has been applied to examine the relationship between flowering time and latitude in angiosperm species, where phylogenetic adjustment reveals significant clinal variation after controlling for shared ancestry and estimating \lambda to quantify signal strength in residuals. Key advantages of PGLS include its ability to handle phylogenies that are not ultrametric, accommodating uneven branch lengths and incomplete sampling through the flexible covariance structure. Extensions to multivariate PGLS further allow simultaneous analysis of multiple response traits, incorporating high-dimensional data such as morphological shapes via distance-based approaches.

Simulation-based methods

Simulation-based methods in phylogenetic comparative analysis involve generating synthetic datasets under specified null evolutionary models on a given phylogeny to empirically evaluate hypotheses about evolution. These approaches, often employing techniques, simulate data to create a of a chosen , allowing researchers to assess whether observed patterns deviate significantly from expectations under models like (BM) or the Ornstein-Uhlenbeck () process. By comparing the observed statistic to this simulated distribution, p-values are derived without relying on parametric assumptions, making the method flexible for complex scenarios where analytical solutions are unavailable. The core procedure entails simulating multiple datasets (typically 1,000 or more) by forward-evolving traits along the phylogeny using the null model, computing the test statistic for each simulated dataset, and then positioning the observed value within the resulting empirical distribution. For instance, under a BM model, trait values are generated by accumulating random increments proportional to branch lengths, with the rate parameter σ² often estimated from the observed data to ensure fair comparison. Similarly, for OU processes, simulations incorporate mean-reverting dynamics with parameters like the optimum θ and strength α fitted from data. This simulation framework enables distribution-free inference, particularly useful for testing associations between traits or deviations from neutrality. Key steps include parameter estimation from the empirical data (e.g., maximum likelihood for σ² in ), repeated simulation of tip traits, and calculation of the test statistic, such as a between two traits. For discrete traits, Brownian threshold tests model binary or multistate characters as thresholds on an underlying continuous evolving under , simulating the liability and applying thresholds to derive observed states for comparison. These tests assess phylogenetic clustering beyond chance, with determined by the proportion of simulated datasets exceeding the observed clustering metric. Typically, 1,000–10,000 iterations suffice for robust estimation, balancing computational cost and precision. A representative example is testing for phylogenetic signal in morphological , such as body size or size proxies, using simulations under to evaluate whether observed variance is more structured by phylogeny than expected by chance. In analyses of datasets, researchers simulate on phylogenies, compute Blomberg's for each replicate, and compare it to the empirical ; significant deviations indicate stronger phylogenetic conservatism, as seen in studies revealing labile behavioral but conserved morphological ones across lineages. This approach has illuminated how size in may exhibit phylogenetic signal exceeding expectations, informing macroevolutionary patterns. Recent advances extend these methods to complex scenarios via Approximate Bayesian Computation (ABC), which uses simulations to approximate posterior distributions for model parameters in non-tractable likelihood cases, such as multivariate trait evolution or regime shifts. ABC in PCMs simulates datasets under candidate models, accepts those yielding close to observed data, and infers parameters like selection strength without full Bayesian integration. Additionally, the 2024 update to the phytools enhanced simulation capabilities, improving support for phylogenetic comparative analyses. Despite their versatility, simulation-based methods are computationally intensive, often requiring hours or days for extensive replicates on large trees, which limits applicability to massive datasets. They also remain sensitive to model misspecification; if the null model (e.g., ) poorly represents the true process, simulated distributions may mislead inference, emphasizing the need for prior model validation through goodness-of-fit simulations.

Advanced Topics

Ancestral state reconstruction

Ancestral state reconstruction is a key application of phylogenetic comparative methods (PCMs) that infers the evolutionary states of traits at internal nodes of a , enabling insights into historical trait evolution under specified models. These methods typically employ maximum parsimony, maximum likelihood (ML), or Bayesian frameworks to estimate ancestral values, accounting for phylogenetic relationships and evolutionary processes such as branching patterns and branch lengths. Maximum parsimony minimizes the total amount of evolutionary change required to explain observed tip states, while ML and Bayesian approaches incorporate probabilistic models of trait evolution to compute likelihoods or posterior probabilities of ancestral states. For continuous traits, squared-change parsimony provides a parsimony-based method to reconstruct ancestral states by minimizing the sum of squared changes along branches, which is mathematically equivalent to ML estimation under a Brownian motion (BM) model of evolution. Under the BM model, the ML estimate of the ancestral state at an internal node i is given by the weighted average of the states of its descendant tips, where the weights w_{ij} for each descendant j are proportional to the inverse of the branch length from the node to the tip: \hat{x}_i = \frac{\sum_{j} w_{ij} x_j}{\sum_{j} w_{ij}}, \quad w_{ij} = \frac{1}{t_{ij}}, with t_{ij} as the total path length from node i to tip j; the rate of evolution \sigma^2 cancels out in the weights. This approach assumes gradual, random diffusion of the trait and performs well for traits evolving neutrally but can underestimate uncertainty in deep divergences. For discrete traits, the Mk model, a continuous-time Markov chain framework, is widely used to estimate transition rates between states and reconstruct ancestral states via ML. Introduced by Pagel (1994), the model parameterizes a rate matrix Q where off-diagonal elements represent transition rates between states (e.g., equal rates for symmetric evolution or unequal rates for directional biases), and the likelihood of ancestral states at a node is computed by integrating over all possible histories consistent with the tip data and phylogeny. The diagonal elements of Q ensure row sums of zero, modeling the instantaneous rate of change, with ancestral probabilities derived from the pruned likelihoods at each node using Felsenstein's pruning algorithm. Bayesian extensions enhance ancestral state reconstruction by incorporating prior distributions on parameters and using (MCMC) sampling to explore posterior distributions of ancestral states, thereby quantifying uncertainty. Software like MrBayes implements this for discrete traits by jointly estimating phylogeny, substitution models, and ancestral states under a hierarchical Bayesian framework, often using the model with priors on transition rates. For greater resolution of evolutionary histories, character mapping (SCM) samples character histories proportional to their posterior probabilities, allowing of multiple plausible mappings rather than point estimates; Bollback (2006) developed SIMMAP for this purpose, enabling probabilistic of transition timings and counts along branches. An illustrative application involves reconstructing preferences in cetaceans, where discrete models (extended to account for rate heterogeneity akin to regime shifts in continuous processes) have inferred multiple transitions from terrestrial to fully states across lineages. Despite these advances, ancestral state reconstruction faces limitations, including a systematic toward present-day tip states in and ML methods under symmetric models, which can distort inferences for rapidly evolving or asymmetrically transitioning traits. Handling regime shifts—abrupt changes in evolutionary dynamics—requires extended models like those allowing time-heterogeneous rates, but these increase computational demands and risk without sufficient data.

Phylogenetic signal and tests

Phylogenetic signal refers to the tendency for closely related to exhibit greater similarity in their than would be expected by chance, reflecting the influence of shared evolutionary history on evolution. This concept is central to phylogenetic comparative methods, as it quantifies the degree to which phylogenetic relatedness structures variation in phenotypic data, often violating assumptions of independence in non-phylogenetic analyses. Detecting and measuring phylogenetic signal is essential for selecting appropriate evolutionary models and interpreting comparative results accurately. One widely used measure is Blomberg's K, introduced as a diagnostic for the strength of phylogenetic signal relative to a (BM) model of . Under BM, K = 1 indicates that observed trait variance matches expectations from phylogenetic drift alone; K > 1 suggests stronger signal than BM (e.g., due to ), while K < 1 implies weaker signal, potentially from or high lability. The statistic is computed as: K = \frac{\text{observed MSO} / \text{expected MSO under BM}}{\text{observed MSE} / \text{expected MSE under BM}} where MSO is the mean squared error among tip values relative to the phylogenetic mean (capturing among-species variance), and MSE is the mean squared error of phylogenetic independent contrasts (capturing within-clade variance). Significance is typically assessed via randomization: tip labels are permuted 1,000 times to generate a null distribution of K under no signal, with p-values derived from the proportion of simulated K values exceeding the observed K. Another prominent measure is Pagel's λ, a branch-length scaling parameter that transforms the phylogenetic to model deviations from . λ ranges from 0 (no phylogenetic signal, equivalent to a star phylogeny) to 1 (full signal, where covariances scale with shared branch lengths); intermediate values indicate partial signal, often due to processes like Ornstein-Uhlenbeck evolution. Fitting λ via maximum likelihood allows hypothesis testing through likelihood ratio comparisons against null models (e.g., λ = 0 or λ = 1), providing a model-based assessment of signal strength. For example, in a study of Iberian freshwater fishes, critical swimming speed (U_crit) exhibited moderate phylogenetic signal with Blomberg's K = 0.415 (p = 0.006), indicating some conservatism beyond BM expectations, though Pagel's λ was non-significant, suggesting additional ecological influences on . Low K values in such traits often highlight adaptive driven by demands, like fast-flowing streams. Advances in measuring phylogenetic signal have extended to multivariate traits, particularly high-dimensional data like geometric morphometrics. Adams (2014) generalized Blomberg's K to a multivariate form (K_mult), which decomposes covariances into phylogenetic and residual components using Procrustes-aligned coordinates, enabling signal detection in shape data where univariate approaches fail. Recent extensions, as of 2025, incorporate phylogenomic datasets to evaluate signal in genomic-scale traits, such as profiles, using unified indices that handle continuous, discrete, and mixed data types for improved robustness in large-scale evolutionary analyses. These methods often rely on simulations for estimation to account for complex covariance structures in phylogenomic trees.

Applications

Studying trait evolution

Phylogenetic comparative methods (PCMs) are widely applied to test for correlated evolution between traits, such as those driven by , by accounting for shared ancestry to identify non-independent associations. For instance, analyses of primary reproductive traits in seed beetles have revealed correlated changes between male and female characteristics, suggesting influenced by sexual selection pressures. PCMs also enable detection of rate heterogeneity in trait evolution across clades, where evolutionary rates vary due to ecological or genetic factors, often modeled using extensions of or Ornstein-Uhlenbeck () processes that incorporate variable rates along phylogenetic branches. A prominent case study involves the of lizards, where OU models applied to ecomorphological traits—such as limb length and body size—demonstrate of similar forms on different Caribbean islands, highlighting how phylogenetic constraints and selection shape parallel adaptations to habitat types. In mammals, phylogenetic generalized least squares (PGLS) analyses of life-history trade-offs, such as the relationship between gestation length, body mass, and lifespan, reveal how these traits covary under evolutionary constraints, with larger species exhibiting longer gestations but adjusted reproductive strategies to balance survival and . PCMs facilitate evolutionary inferences by detecting shifts in trait optima, as in the l1ou method, which uses a lasso-penalized approach under the model to identify branches where selective regimes change, allowing reconstruction of adaptive transitions in continuous traits across phylogenies. Disparity-through-time analyses further quantify how morphological variation accumulates or stabilizes over phylogenetic history, often revealing early bursts of disparity in radiating clades followed by stabilization, as seen in cetacean body size evolution where initial rapid diversification outpaced lineage accumulation. Recent applications extend PCMs to phylogenomic data for studying trait evolution in pathogens, such as in viruses where high-resolution phylogenies enable of rates to host environments; for example, analyses of leek yellow stripe virus (LYSV) coevolution with plants in 2023 demonstrated host-specific shifts in viral traits like transmission efficiency, underscoring rapid adaptive responses in lineages. These studies yield insights into by pinpointing selective optima and reveal phylogenetic constraints limiting trait lability.

Macroecological patterns

Phylogenetic comparative methods (PCMs) have been instrumental in analyzing macroecological patterns, such as latitudinal diversity gradients, by incorporating phylogenetic structure to disentangle evolutionary history from ecological processes. For instance, studies using phylogenetic diversity metrics have revealed contrasting patterns in species richness and evolutionary distinctiveness across latitudes, with higher phylogenetic diversity at higher latitudes for woody communities due to patterns in community structure. PCMs correct for phylogenetic non-independence in analyses of diversification, enabling assessments of how factors like island size and isolation influence rates beyond simple taxonomic counts. In community phylogenetics, these methods quantify the role of evolutionary relatedness in structuring local assemblages, revealing how phylogenetic clustering or overdispersion reflects environmental filtering or competitive interactions across biogeographic scales. Case studies illustrate the application of PCMs to global patterns, particularly in avian body size clines. Using phylogenetic generalized least squares (PGLS), analyses of Australian passerines across 82 demonstrated that body size tracks climatic variation, with larger sizes in cooler regions conforming to after accounting for phylogenetic relationships, highlighting shifts linked to warming temperatures. In island biogeography, PCMs integrate trait-phylogeny links to model and ; for example, comparative analyses of island floras have shown that phylogenetic assessments of speciation modes outperform taxonomy-based metrics in predicting equilibria, as seen in studies of oceanic archipelagos where conservatism influences . Insights from PCMs include the use of phylogenetic dispersion metrics like mean pairwise distance () and mean nearest distance (MNTD) to infer assembly rules, where captures deep-time clustering indicative of filtering, while MNTD highlights recent divergences from interactions in diverse ecosystems. Recent applications, such as 2024 studies on , employ phylogenetic analyses to forecast shifts in phylogenetic , predicting homogenization of floristic regions under future warming scenarios across global distributions. These metrics underscore how evolutionary legacies shape macroecological responses to environmental gradients. Challenges in applying PCMs to macroecology arise from scaling phylogenetic analyses from species-level s to , where integrating multi-scale often requires assumptions about unobserved processes that may diversification estimates. Incomplete sampling in datasets exacerbates these issues, as nonrandom taxa can distort correlations and phylogenetic signal detection, particularly in underrepresented tropical clades, necessitating simulation-based corrections to ensure robust inferences. A representative example is the analysis of plant functional traits, such as leaf area and wood density, comparing tropical and temperate floras; models across angiosperm phylogenies reveal stronger phylogenetic conservatism in tropical traits adapted to stable climates, contrasting with greater lability in temperate zones, informing predictions of community responses to shifting biomes.

Implementation and Software

Key software packages

Phylogenetic comparative methods (PCMs) are predominantly implemented through open-source software packages, with the R programming language offering the most comprehensive ecosystem due to its flexibility and integration with statistical tools. Key packages in R include phytools, which supports a wide array of analyses such as phylogenetically independent contrasts (PIC), phylogenetic generalized least squares (PGLS), ancestral state reconstruction, simulations under Ornstein-Uhlenbeck (OU) models, and phylogenetic signal tests via functions like phylosig, with version 2.5-2 released in September 2025 and actively maintained. The ape package provides essential functions for phylogenetic tree manipulation, reading/writing tree formats, and basic comparative computations, serving as a foundational tool for PCM workflows. Complementing these, picante enables analyses of phylogenetic diversity in ecological communities, including metrics like Faith's phylogenetic diversity and null model tests for community structure. Additionally, geiger facilitates model fitting to phylogenetic trees, such as branching time simulations and tests for phylogenetic signal (e.g., Moran's I), with version 2.0.11 expanding support for large datasets and macroevolutionary models. Beyond R, software in other languages addresses specialized needs. BayesTraits, a standalone program, implements Bayesian approaches for discrete and continuous trait evolution, including reversible-jump MCMC for model comparison and ancestral state reconstruction, and is actively updated (version 5.0.3 as of October 2025). Mesquite offers a for phylogenetic analysis, supporting comparative data management, tree , and modules for PCMs like ancestral states, making it suitable for interactive exploration. In , emerging tools as of 2025 include Profylo, an open-source package for phylogenetic profile comparison and co-evolution analysis, alongside libraries like TreeSwift for scalable tree manipulation, reflecting growing integration of PCMs with Python's ecosystem. These packages were selected based on their open-source nature, active maintenance through repositories like CRAN or , and widespread adoption in peer-reviewed studies, ensuring reliability for PCM applications. For instance, a PGLS in using phytools and might involve loading a and , then applying the : first, prepare with trait ~ 1 for intercept-only models; fit via pgls([trait](/page/Trait) ~ predictor, [data](/page/Data) = comparative_data, [tree](/page/Tree) = phy_tree); and summarize results with standard errors accounting for phylogenetic . Recent updates as of 2025 emphasize scalability, with R packages like rotl integrating directly with the for accessing large, synthesized phylogenies in PCM analyses.

Practical considerations

Applying phylogenetic comparative methods (PCMs) requires careful verification of key assumptions to ensure reliable inferences about trait evolution. A fundamental assumption is the accuracy of the phylogenetic tree, including its topology and branch lengths, as errors in tree reconstruction can propagate biases into downstream analyses. For instance, polytomies or unresolved branches may underestimate evolutionary variance, leading to inflated type I error rates in tests of trait correlations. Model adequacy must also be assessed, often using information criteria such as Akaike's Information Criterion (AIC) to compare simpler models like (BM) against more complex ones like (OU), particularly since small datasets (e.g., fewer than 50 taxa) can erroneously favor OU due to sampling variance at tree tips. Handling is another critical step; methods like phylogenetic imputation via multivariate models (e.g., Rphylopars) estimate absent trait values by leveraging phylogenetic , preserving correlations better than listwise deletion, which reduces statistical power. Diagnostics play a vital role in validating PCM results and identifying potential issues. Residual plots from phylogenetic generalized least squares (PGLS) regressions should be examined for patterns of autocorrelation not captured by the phylogeny, while phylogenetic signal tests—such as Blomberg's or Pagel's λ—quantify the degree to which traits covary with phylogenetic relatedness, with values near 1 indicating strong signal under BM-like . Sensitivity analyses, facilitated by tools like the sensiPhy, assess robustness to branch length estimation by resampling multiple trees (e.g., from Bayesian posteriors) and evaluating how influential taxa or clades affect parameter estimates. These diagnostics reveal, for example, that short branch lengths in molecular phylogenies can overestimate evolutionary rates compared to fossil-calibrated trees, which incorporate deeper temporal information but may introduce calibration biases from sparse fossil records. Despite their power, PCMs face notable limitations that users must navigate. Computational demands escalate with large phylogenies (e.g., thousands of taxa), as likelihood calculations in multivariate models require substantial memory and time, often necessitating parallel computing or approximations like sub-tree pruning for scalability. Biases arise from tree type: molecular trees, estimated from sequence data, tend to compress deep branches and inflate recent ones, potentially underestimating ancestral trait reconstruction accuracy, whereas fossil-calibrated trees mitigate this but suffer from uncertainties in fossil placement and minimum age constraints. Additionally, incomplete sampling in phylogenies can lead to low statistical power, especially for detecting weak signals in trait evolution. Best practices enhance the reliability of PCM applications. Analyses should begin with phylogenetic signal tests to confirm non-random trait distribution before proceeding to model fitting, followed by AIC-based selection to avoid . Sensitivity to phylogenetic uncertainty should be reported, including how results vary across alternative , and full phylogenetic details—such as tree source, calibration method, and branch length units—must be disclosed in publications to enable . For small phylogenies (e.g., <20 taxa), power analyses via simulations are essential, as they often yield low detection rates for evolutionary shifts; in such cases, or Bayesian approaches can quantify uncertainty more robustly than frequentist tests. As of 2025, emerging issues in PCMs include managing incomplete phylogenies from metagenomic datasets, where microbial communities yield fragmented trees due to and undersampling; approaches like augmentation improve resolution by integrating tailored orthologs to infer deeper relationships. Ethical considerations in data sharing are also prominent, emphasizing for genomic data underlying phylogenies, equitable access to avoid exacerbating disparities, and safeguards against re-identification in shared databases. For example, in analyses of small phylogenies like those of , low power can be troubleshot by incorporating intraspecific variation or proxies to bolster sample size, though this requires validating assumptions of across populations.

References

  1. [1]
    Phylogenetic comparative methods - ScienceDirect.com
    May 8, 2017 · Phylogenetic comparative methods (PCMs) are statistical methods for studying organismal evolution, using species relatedness and trait values. ...
  2. [2]
    Phylogenies and the Comparative Method | The American Naturalist
    The phylogeny specifies a set of contrasts among species, contrasts that are statistically independent and can be used in regression or correlation studies.
  3. [3]
    Shedding light on the 'dark side' of phylogenetic comparative methods
    Summary. Phylogenetic comparative methods are becoming increasingly popular for investigating evolutionary patterns and processes.
  4. [4]
    Phylogenies and the Comparative Method - jstor
    9.-A less symmetrical phylogeny. The independent contrasts for this phylogeny are given in table 1. TABLE 1. THE FOUR CONTRASTS EXTRACTED FROM THE PHYLOGENY.
  5. [5]
    [PDF] Phylogenetic Comparative Methods - Harmon Lab
    These papers raise substantive critiques that are important to address when applying the methods described in this chapter to empirical data. In this.
  6. [6]
    Comparative anatomy | Definition, Examples, & Facts - Britannica
    Cuvier also rejected the 18th-century notion that the members of the animal kingdom are arranged in a single linear series from the simplest up to humans.
  7. [7]
    [PDF] Karl Pearson and Statistics: The Social Origins of Scientific Innovation
    Karl Pearson (1857-1936) is often considered to be the father of the modern discipline of statistics, which emerged from his work in mathematical biology or.
  8. [8]
    Bayesian models for comparative analysis integrating phylogenetic ...
    Jun 28, 2012 · We have shown that Bayesian methods for phylogenetic comparative analysis are easy to implement in the BUGS language, often only requiring ...Missing: post- | Show results with:post-
  9. [9]
    Phylogenomic comparative methods: Accurate evolutionary ... - PNAS
    Phylogenetic comparative methods allow for the study of trait evolution between species by accounting for their shared evolutionary history.Missing: key | Show results with:key
  10. [10]
    Reading trees: A quick review - Understanding Evolution
    A phylogeny, or evolutionary tree, represents the evolutionary relationships among a set of organisms or groups of organisms, called taxa (singular: taxon).Missing: topology polytomies<|separator|>
  11. [11]
    A Metric on Phylogenetic Tree Shapes - PMC - PubMed Central
    A multifurcation, or a polytomy, is a node with more than two children, and its size is its number of children ( ). Naturally, a rooted phylogeny defines a ( ...
  12. [12]
    RAxML-III: a fast program for maximum likelihood-based inference of ...
    ... MrBayes outperforms all other analyzed phylogeny programs in terms of speed and tree quality. MrBayes is a program for bayesian analysis of phylogenetic trees.
  13. [13]
    Non-ultrametric phylogenetic trees shed new light on Neanderthal ...
    Jun 10, 2023 · Ultrametric spaces are widely used to depict evolutionary times in phylogenetic trees since they assume that every population/species is ...
  14. [14]
    A comparative method for both discrete and continuous characters ...
    Dec 15, 2011 · The threshold model developed by Sewall Wright in 1934 can be used to model the evolution of two-state discrete characters along a phylogeny.
  15. [15]
    Phylogenetic Comparative Methods for Evaluating the Evolutionary ...
    Phylogenetic comparative methods offer a suite of tools for studying trait evolution. However, most models inherently assume fixed trait values within species.
  16. [16]
    Handling missing values in trait data - Johnson - Wiley Online Library
    Oct 18, 2020 · The standard practice in comparative studies that use imputation is to impute values using only the traits and, where relevant, the phylogeny.
  17. [17]
    The evolution of mammalian brain size | Science Advances
    Apr 28, 2021 · Our results reveal that the largest-brained mammals achieved large relative brain sizes by highly divergent paths.
  18. [18]
    Inferring the mammal tree: Species-level sets of phylogenies for ...
    Dec 4, 2019 · We compare our phylogenetic trees to previous estimates of mammal-wide phylogeny and divergence times, finding that (1) node ages are broadly ...
  19. [19]
    Shedding light on the 'dark side' of phylogenetic comparative methods
    Jun 13, 2016 · Phylogenetic comparative methods (PCMs) were initially developed in the 1980s to deal with the statistical non-independence of species in ...Missing: Cuvier Pearson
  20. [20]
    Unsolved Challenge to Phylogenetic Correlation Tests for ...
    Our Methods Do Not Correct Pseudoreplication. A primary motivation for developing tree-based quantitative methods for detecting correlated trait evolution is ...Model Failure and... · Vigilance Is Not Enough · The Curvature of Biodiversity...
  21. [21]
    Wing pointedness associated with migratory distance in common ...
    Apr 19, 2010 · ... phylogenetic comparative methods to test the prediction that ... Wing length was significantly correlated with migration scores for ...
  22. [22]
  23. [23]
    Bayesian Analyses of Comparative Data with the Ornstein ...
    Phylogenetic comparative methods (PCMs) aim to analyze data sets of species traits in a phylogenetic framework. They are used for various purposes, including ...Missing: post- | Show results with:post-
  24. [24]
    Phylogenies and the Comparative Method | The American Naturalist
    The phylogeny specifies a set of contrasts among species, contrasts that are statistically independent and can be used in regression or correlation studies.
  25. [25]
    Metabolic costs of brain size evolution | Biology Letters - Journals
    Thus, we conducted both an analysis of family means and an analysis using phylogenetically independent contrasts, as proposed by Felsenstein (1985). Contrasts ...
  26. [26]
    A Primer on Phylogenetic Generalised Least Squares - SpringerLink
    A Primer on Phylogenetic Generalised Least Squares. Chapter. pp 105–130; Cite this chapter. Download book PDF · Download book EPUB · Modern Phylogenetic ...
  27. [27]
  28. [28]
    Phylogenetic conservatism and trait correlates of spring ...
    Jul 22, 2017 · Our study confirms that both spring phenology (both leaf-out and flowering phenology) and their responses to spring temperature are constrained by phylogeny.
  29. [29]
    Assessing the Goodness of Fit of Phylogenetic Comparative Methods
    Phylogenetic comparative methods (PCMs) have been applied widely in analyzing data from related species but their fit to data is rarely assessed. Question. Can ...
  30. [30]
    Monte Carlo simulation in phylogenies: an application to test the ...
    Monte Carlo simulation in phylogenies tests tree-reconstruction methods and evolutionary models by estimating probability distributions for parameters.
  31. [31]
    Testing for phylogenetic signal in comparative data: behavioral traits ...
    We present new methods for continuous-valued characters that can be implemented with either phylogenetically independent contrasts or generalized least-squares ...
  32. [32]
    A Comparative Method for Both Discrete and Continuous Characters ...
    The threshold model developed by Sewall Wright in 1934 can be used to model the evolution of two-state discrete characters along a phylogeny.
  33. [33]
    Detecting Phenotypic Selection by Approximate Bayesian ...
    This chapter discusses the fundamental structure and advantages of the approximate Bayesian computation (ABC) algorithm in phylogenetic comparative methods ...
  34. [34]
    [PDF] phytools 2.0: an updated R ecosystem for phylogenetic comparative ...
    Jan 5, 2024 · phytools is a diverse contributed R library now consisting of hundreds of different functions covering a variety of methods and purposes in ...
  35. [35]
    Ancestral Reconstruction | PLOS Computational Biology
    Jul 12, 2016 · Ancestral reconstruction can be thought of as the direct result of applying a hypothetical model of evolution to a given phylogeny. When the ...Missing: seminal | Show results with:seminal
  36. [36]
    Bayesian Estimation of Ancestral Character States on Phylogenies
    We use a continuous-time Markov model (Pagel, 1994, 1997; Lewis, 2001), mathematically equivalent to models of gene-sequence evolution, but not limited to ...
  37. [37]
  38. [38]
    SIMMAP: Stochastic Mapping of Discrete Traits on Phylogenies
    Feb 23, 2006 · SIMMAP has been developed to implement stochastic character mapping that is useful to both molecular evolutionists, systematists, and bioinformaticians.
  39. [39]
    [PDF] A phylogenetic blueprint for a modern whale
    Oct 1, 2012 · We mapped the evolution of habitat, freshwater versus marine, onto the crown. Cetacea section of our composite tree (Fig. 9). Each family of ...<|separator|>
  40. [40]
    Baleen Whale Phylogeny and a Past Extensive Radiation Event ...
    We reconstructed the phylogenetic relationships of baleen whales by characterizing 36 informative SINE loci.
  41. [41]
    Limitations of Ancestral State Reconstruction and the Evolution of ...
    Ancestral state reconstructions were performed using maximum parsimony and maximum likelihood on a posterior tree sample as well as two fully Bayesian methods.Missing: day | Show results with:day
  42. [42]
    Accuracy of ancestral state reconstruction for non-neutral traits
    May 6, 2020 · In this paper, we conduct a simulation study to assess the accuracy of common methods of ancestral state reconstruction. We exploit the ...Missing: seminal | Show results with:seminal
  43. [43]
    How to measure and test phylogenetic signal
    Blomberg's K. Blomberg's K expresses the strength of phylogenetic signal as the ratio of the mean squared error of the tip data (MSE0) measured from the ...
  44. [44]
    Generalized K Statistic for Estimating Phylogenetic Signal from ...
    In this article, I describe a generalization of the K statistic of Blomberg et al. that is useful for quantifying and evaluating phylogenetic signal in highly ...Abstract · Multivariate Generalization of... · Statistical Performance of... · DiscussionMissing: autocorrelation | Show results with:autocorrelation
  45. [45]
    How to Define, Use, and Interpret Pagel's λ$$ \lambda $$ (Lambda ...
    Apr 9, 2025 · Pagel's λ $$ \lambda $$ (lambda) is a useful tool in ecology and evolution for describing trait evolution, imputing missing species' data, ...Introduction · Defining Pagel's · Phylogenetic Imputation of... · Hierarchical Models
  46. [46]
    A comparison of metrics for estimating phylogenetic signal under ...
    Aug 2, 2012 · Here, two of these model-based metrics (Blomberg's K-statistics and Pagel's λ) are compared with three statistical approaches.
  47. [47]
    [PDF] Phylogenetic signal and evolutionary relationships among traits of ...
    Jan 10, 2022 · Specifically, species- specific critical swimming speed (Ucrit), a measure of a species' swim- ming performance, was estimated from a recently ...
  48. [48]
    Key factors explaining critical swimming speed in freshwater fish
    Nov 3, 2020 · For example, a previous study showed that cyprinids living in fast flowing habitats showed higher Ucrit values compared to fish species ...
  49. [49]
    A Unified Method for Detecting Phylogenetic Signals in Continuous ...
    Mar 20, 2025 · We developed a new index (the M statistic) to detect phylogenetic signals for continuous traits, discrete traits, and multiple trait combinations.
  50. [50]
    Exploring Phylogenetic Signal in Multivariate Phenotypes by ...
    Blomberg's K is a common statistic to quantify phylogenetic signal (Blomberg et al., 2003). ... All simulations were performed in R, using routines written ...
  51. [51]
    Correlated evolution between male and female primary reproductive ...
    Nov 10, 2010 · Here, we employ a phylogenetic comparative approach to assess correlated evolution between primary reproductive characters in males and those in ...
  52. [52]
    The impact of rate heterogeneity on inference of phylogenetic ...
    Rates of trait evolution are known to vary across phylogenies; however, standard evolutionary models assume a homogeneous process of trait change.
  53. [53]
    The evolution of gestation length in eutherian mammals - PMC - NIH
    Oct 30, 2024 · We found that gestation length substantially differed in both whether and how strongly it was associated with body mass and lifespan across mammals.
  54. [54]
    Fast and accurate detection of evolutionary shifts in Ornstein ...
    Jan 7, 2016 · Hansen (1997) used an Ornstein–Uhlenbeck (OU) process to model the macroevolution of a phenotype subject to selection pressure towards an ...
  55. [55]
    Host-specific adaptation drove the coevolution of leek yellow stripe ...
    Sep 14, 2023 · We show that the evolutionary context of LYSV has been shaped by the host adaptation of the virus during its coevolution with Allium plants.
  56. [56]
    A Guided Tour of Phylogenetic Comparative Methods for Studying ...
    Nov 4, 2024 · Phylogenetic comparative methods are important tools in biology, providing insights into the way traits evolve. There are many technical ...
  57. [57]
    Contrasting latitudinal patterns in phylogenetic diversity between ...
    Apr 23, 2019 · The latitudinal diversity gradient (LDG), which shows a pattern of decreasing species richness from the equator towards the poles is one of the ...
  58. [58]
    Large-scale phylogenetic analyses reveal the causes of high ... - NIH
    A negative latitudinal gradient in species richness (i.e. higher tropical diversity) is one of the most well-known patterns in biology [1] and has been ...
  59. [59]
    Using phylogenetic data for island biogeography analyses
    DAISIE is an island biogeography model that can estimate rates of colonisation, speciation and extinction from island communit y phylogenies.
  60. [60]
    Australian songbird body size tracks climate variation: 82 species ...
    Nov 27, 2019 · We used phylogenetic generalized least squares (PGLS) ... Shifting latitudinal clines in avian body size correlate with global warming in ...
  61. [61]
    Shifting latitudinal clines in avian body size correlate with global ...
    Four species showed significant decreases in body size (1.8–3.6% of wing length) and a shift in latitudinal cline over that period, and a meta-analysis ...Missing: PGLS | Show results with:PGLS
  62. [62]
    A comparative analysis of island floras challenges taxonomy‐based ...
    Importantly, comparing phylogenetic assessments of the modes of speciation with taxonomy‐based measures shows that widely used island biogeography approaches ...
  63. [63]
    Community phylogenetic structure reveals the imprint of dispersal ...
    Aug 12, 2019 · The MPD index is a basal metric which is sensitive to deep phylogenetic structures, whereas the MNTD index is a terminal metric which is ...Missing: phylogenetics | Show results with:phylogenetics
  64. [64]
    For common community phylogenetic analyses, go ahead and use ...
    Jun 21, 2019 · PD = Faith's PD, MPD = mean pairwise distance, MNTD = mean pairwise distance between the closest relatives. Within the model, the response ...
  65. [65]
    Climate change alters the future of natural floristic regions of deep ...
    Nov 2, 2024 · Our analysis reveals declines in phylogenetic beta diversity for years 2040 to 2100, leading to a future homogenization of biogeographic regions.
  66. [66]
    Realistic scenarios of missing taxa in phylogenetic comparative ...
    Oct 11, 2019 · For example, trees with up to 128 tips sometimes do not allow the correct selection of an OUM model over OU (Beaulieu et al., 2012). The ...
  67. [67]
    Nonrandom Variation in Within-Species Sample Size and Missing ...
    Jun 28, 2011 · Such phylogenetic comparative methods have been extensively used to make inferences about correlated trait evolution (Martins and Hansen ...THE SOURCES OF... · CONSEQUENCES OF... · RECOMMENDATIONS FOR...Missing: challenges macroecology scaling
  68. [68]
    Assessing the impact of incomplete species sampling on estimates ...
    Mar 16, 2020 · We compared our approach with a phylogenetic model that does not incorporate incomplete species sampling and to three fossil-based alternatives ...Per-Taxon Rates And The... · Extant Species Sampling · Results
  69. [69]
    Global patterns of plant functional traits and their relationships to ...
    Sep 13, 2024 · Plant functional traits (FTs) determine growth, reproduction and survival strategies of plants adapted to their growth environment.
  70. [70]
    Mechanisms driving plant functional trait variation in a tropical forest
    Mar 13, 2021 · Plant functional trait variation in tropical forests results from taxonomic differences in phylogeny and associated genetic differences, as well as, phenotypic ...
  71. [71]
    phytools 2.0: an updated R ecosystem for phylogenetic comparative ...
    Jan 5, 2024 · Phylogenetic comparative methods comprise the general endeavor of using an estimated phylogenetic tree (or set of trees) to make secondary ...<|control11|><|separator|>
  72. [72]
    [PDF] phytools.pdf - CRAN
    Sep 19, 2025 · Revell, L. J. (2024) phytools 2.0: an updated R ecosystem for phylogenetic comparative methods. (and other things). PeerJ, 12, e16505. add ...
  73. [73]
    Picante: R tools for integrating phylogenies and ecology
    Summary: Picante is a software package that provides a comprehensive set of tools for analyzing the phylogenetic and trait diversity of ecological communities.Abstract · INTRODUCTION · DESCRIPTION
  74. [74]
    geiger v2.0: an expanded suite of methods for fitting ...
    Geiger v2.0 is a complete overhaul of the R package geiger, designed to scale up comparative methods to large datasets and address computational issues.Abstract · INTRODUCTION · METHODS · CONCLUSION
  75. [75]
    [PDF] BayesTraits V4.0.0 Manual updated October 2023 Andrew Meade ...
    Traditionally comparative methods assume trait values are known without error but traits often show within species variation. BayesTraits allows samples of ...
  76. [76]
    Mesquite Project
    Mesquite is designed to help phylogenetic and evolutionary biologists organize and analyze comparative data about organisms.Download and InstallationMesquite ManualPhylogenetic treesWhat Mesquite DoesAncestral States
  77. [77]
    A Python Package for Phylogenetic Profile Comparison and Analysis
    Oct 29, 2025 · We present Profylo, a phylogenetic profiling toolkit made available as an open-source Python 3.0 package. Profylo implements seven methods for ...
  78. [78]
    TreeSwift: A massively scalable Python tree package - ScienceDirect
    We present TreeSwift, a user-friendly and massively scalable Python package for traversing and manipulating trees that is ideal for algorithms performed on ...Missing: comparative | Show results with:comparative<|separator|>
  79. [79]
    OpenTree: A Python Package for Accessing and Analyzing Data ...
    Open Tree of Life provides web-service application programming interfaces (APIs) to make the tree estimate, unified taxonomy, and input phylogenetic data ...
  80. [80]
    Phylogenetic Methods Meet Deep Learning - PubMed
    Sep 19, 2025 · Deep learning (DL) has been widely used in various scientific fields, but its integration into phylogenetics has been slower, primarily due ...Missing: hybrids comparative 2020-2025
  81. [81]
    Rphylopars: fast multivariate phylogenetic comparative methods for ...
    Jun 26, 2016 · Dealing with missing data in comparative studies. A frequently encountered problem in comparative studies is the difficulty (or impossibility) ...
  82. [82]
    sensiPhy: An r‐package for sensitivity analysis in phylogenetic ...
    Feb 26, 2018 · We present sensiPhy, an r-package, to easily and rapidly perform sensitivity analysis for phylogenetic comparative methods. sensiPhy allows ...
  83. [83]
    Potential for bias and low precision in molecular divergence time ...
    These studies have focused on both fossil calibration and molecular modeling and have contributed a wealth of fascinating insight and useful recommendations but ...Missing: PCM | Show results with:PCM
  84. [84]
    Comparative metagenomics indicates metabolic niche differentiation ...
    Jun 17, 2025 · A phylogenetic reconstruction using all bacterial gspp from the sediment and water column revealed significant differences in microbial ...Metagenome Analyses Using... · Metagenomic Assembly And... · Discussion
  85. [85]
    WHO releases new principles for ethical human genomic data ...
    Nov 20, 2024 · The World Health Organization (WHO) has issued a set of principles for the ethical collection, access, use and sharing of human genomic data.