Phylogenetic Tree

The concept of depicting the relationships among organisms as a branching tree has deep roots in the history of biology. Charles Darwin sketched the earliest known diagram resembling a phylogenetic tree in his Notebook B in 1837, famously headed with the words "I think." This simple branching sketch illustrated his emerging idea that species descend with modification from common ancestors. In On the Origin of Species (1859), Darwin included only a single figure—a hypothetical branching diagram showing how ancestral forms might diversify over geological time into numerous descendant species. Although Darwin's diagram was abstract and did not name actual taxa, it established the conceptual foundation for all subsequent phylogenetic trees.

The German zoologist Ernst Haeckel was the first to publish explicitly named phylogenetic trees based on Darwinian principles. In his 1866 work Generelle Morphologie der Organismen, Haeckel presented elaborate tree diagrams—which he termed "Stammbäume" (genealogical trees)—depicting the relationships among all major groups of life. These included his iconic oak-tree-shaped diagrams showing three kingdoms (Plantae, Protista, Animalia) diverging from a common root. Haeckel also coined the term "phylogeny" (Phylogenie) in this same publication. While many of Haeckel's specific groupings have since been revised, his fundamental approach of using tree diagrams to represent evolutionary history remains central to biology.

The modern methodology for constructing phylogenetic trees was largely shaped by the German entomologist Willi Hennig (1913–1976). In his 1950 work Grundzüge einer Theorie der Phylogenetischen Systematik and the 1966 English revision Phylogenetic Systematics, Hennig articulated the principles that would become known as cladistics. His key contributions include the insistence that only shared derived characters (synapomorphies) provide valid evidence for grouping taxa, the requirement that all recognized groups be monophyletic (clades), and the principle that relationships among species should be interpreted strictly as sister-lineage (clade) relations. Hennig's framework transformed systematics from a largely intuitive discipline into a rigorous, testable methodology. The Willi Hennig Society, founded in his honor, continues to publish the journal Cladistics, which remains a leading venue for phylogenetic research.

A phylogenetic tree consists of several structural elements. Nodes are the branching points: terminal (external) nodes represent the taxa being studied, while internal nodes represent inferred ancestors. The topmost internal node in a rooted tree is the root, symbolizing the most recent common ancestor of all taxa depicted. Branches (also called edges) connect nodes and represent lineages through evolutionary time. Two lineages that diverge from the same internal node are called sister taxa (or sister groups). A node that splits into more than two descendant lineages is called a polytomy, which typically indicates unresolved relationships rather than a true simultaneous divergence event. A basal taxon is one that diverges early from the root and remains unbranched.

Trees may be presented in several formats depending on what information is encoded in branch lengths. A cladogram uses branch lengths that have no quantitative meaning; it shows only the topology (branching order) of relationships. A phylogram has branch lengths proportional to the amount of inferred character change (e.g., nucleotide substitutions). A chronogram (or time-calibrated tree) scales branch lengths to units of geological time, often calibrated using fossil data or molecular clock methods. An ultrametric tree is a special case in which all tips are equidistant from the root, reflecting the assumption that all extant taxa have been evolving for the same amount of time since their last common ancestor.

Phylogenetic trees can be constructed using a range of analytical methods, broadly divided into distance-based and character-based approaches.

Distance-based methods convert a character matrix (e.g., aligned DNA sequences) into a pairwise distance matrix representing evolutionary divergence between each pair of taxa, and then apply clustering algorithms to generate a tree. The most widely used distance method is the Neighbor-Joining (NJ) algorithm, developed by Saitou and Nei in 1987. NJ is computationally efficient and performs well with large datasets, making it a common first-pass approach. However, converting sequence data into distances can result in the loss of phylogenetically informative detail.

Maximum Parsimony (MP), proposed by Farris and Fitch in 1970–1971, seeks the tree that requires the fewest evolutionary changes (character-state transitions) to explain the observed data. It is grounded in the principle of Occam's razor. While conceptually straightforward and requiring no explicit model of sequence evolution, parsimony can be statistically inconsistent under certain conditions—particularly when rates of evolution vary greatly among lineages, a phenomenon known as long-branch attraction (first demonstrated theoretically by Felsenstein in 1978).

Maximum Likelihood (ML), introduced to phylogenetics by Felsenstein in the early 1980s, evaluates trees under an explicit statistical model of sequence evolution (e.g., the General Time Reversible model, GTR). For each candidate tree topology and set of branch lengths, ML calculates the probability (likelihood) that the observed data would have been produced. The tree with the highest likelihood is selected as optimal. ML methods are statistically robust and less susceptible to systematic errors than parsimony, but they are computationally demanding.

Bayesian Inference (BI), introduced to phylogenetics by Rannala and Yang in the 1990s, combines prior probability distributions with the likelihood of the data (using Bayes' theorem) to generate posterior probability distributions for tree topologies and parameters. Markov chain Monte Carlo (MCMC) sampling is used to explore tree space. The tree topology sampled most frequently during the MCMC run is taken as the best estimate. BI has the advantage of providing a natural measure of branch support (posterior probabilities) and can handle complex models efficiently.

For multi-gene datasets, two major integrative strategies exist. The supermatrix (concatenation) method joins aligned sequences of multiple genes end-to-end into a single combined matrix and analyzes it as one. The supertree (coalescence) method first constructs individual gene trees and then integrates them into a single species tree, accounting for potential gene-tree/species-tree discordance caused by processes such as incomplete lineage sorting.

In paleontology, phylogenetic trees are indispensable for understanding the evolutionary relationships of extinct organisms. Because DNA is rarely preserved in fossils, paleontological phylogenetic analyses rely primarily on morphological characters—features of skeletal anatomy, dentition, integument, and other preservable structures. Morphological character matrices are typically analyzed using parsimony or, increasingly, Bayesian methods adapted for discrete morphological data (e.g., the Mk model).

A particularly powerful approach is total-evidence dating (or tip-dating), which integrates morphological data from fossil taxa and molecular data from extant taxa into a single Bayesian analysis. This approach simultaneously infers the tree topology, the placement of fossils, and the absolute timing of divergence events. It has been widely applied to groups such as dinosaurs, early mammals, and Paleozoic invertebrates.

Phylogenetic trees have been transformative for dinosaur systematics. Major reclassifications—such as the revised relationships among theropods showing that birds (Aves) are deeply nested within Maniraptora, or debates about whether Ornithischia and Theropoda form a clade (as proposed in the 2017 "Ornithoscelida" hypothesis by Baron et al.)—are products of cladistic phylogenetic analysis. The placement of newly discovered fossil taxa on existing phylogenetic trees refines our understanding of when and how key adaptations (such as feathers, flight, or herbivory) evolved.

The concept of the molecular clock was proposed by Emile Zuckerkandl and Linus Pauling in the early 1960s (with key publications in 1962 and 1965). They observed that the rate of amino acid substitution in hemoglobin proteins appeared roughly constant over time, suggesting that molecular divergence could be used to estimate the timing of evolutionary splits. While the strict molecular clock assumption has been relaxed in modern analyses (because evolutionary rates vary among lineages), the underlying principle is now fundamental to divergence time estimation. Modern relaxed-clock Bayesian methods allow rates to vary across the tree and use fossil calibration points to convert relative molecular divergence into absolute geological time.

The ambition to construct a comprehensive phylogenetic tree encompassing all of life has driven several major collaborative projects. The Open Tree of Life project published its first draft synthetic tree in 2015 (Hinchliff et al., PNAS), encompassing approximately 2.3 million named species. This tree was assembled by synthesizing published phylogenetic studies with taxonomic data. As of recent updates, the synthetic tree comprises roughly 2.4 million tips, with relationships for approximately 87,000 taxa informed by over 1,200 peer-reviewed phylogenetic studies. The remainder of the tree is scaffolded by taxonomy where no phylogenetic data are available. The Open Tree of Life is freely accessible online and continues to be updated as new studies are incorporated.

Other significant resources include TreeBASE, a repository for published phylogenetic data, and the Tree of Life Web Project (tolweb.org), which aimed to provide information pages for every species and group of organisms.

Phylogenetic trees are hypotheses and are subject to several well-known sources of error and uncertainty. Long-branch attraction (LBA) occurs when rapidly evolving lineages are erroneously grouped together due to convergent substitutions, particularly under parsimony analysis. Gene tree–species tree discordance arises because the evolutionary history of individual genes may differ from that of the species due to incomplete lineage sorting, gene duplication and loss, or horizontal gene transfer. Missing data, especially prevalent in paleontological datasets where preservation is incomplete, can reduce the resolution and accuracy of inferred trees. Model misspecification (using an evolutionary model that poorly fits the data) can also bias results.

To assess the reliability of inferred relationships, several statistical measures are employed. Bootstrap support (introduced by Felsenstein in 1985) resamples characters from the data matrix and measures the proportion of resampled datasets that recover a given clade. Posterior probabilities in Bayesian analyses indicate the probability of a clade given the data and model. Bremer support (decay index) measures how many additional steps are required before a clade is no longer recovered in a parsimony analysis.

Phylogenetic trees have wide-ranging applications far beyond taxonomy. In epidemiology, phylogenetic analysis of pathogen genomes is used to trace the origin, transmission pathways, and evolution of infectious diseases—as demonstrated during the tracking of MRSA, Ebola, and SARS-CoV-2 outbreaks. In conservation biology, phylogenetic diversity metrics help prioritize species and regions for protection, ensuring that the maximum amount of evolutionary history is preserved. The EDGE (Evolutionarily Distinct and Globally Endangered) program uses phylogenetic distinctiveness combined with extinction risk to identify priority species for conservation action. In comparative biology, phylogenetic trees provide the statistical framework for testing hypotheses about trait evolution, adaptation, and correlated changes among characters using methods such as phylogenetic independent contrasts and phylogenetic generalized least squares (PGLS). In biogeography, time-calibrated trees combined with geographic data reveal the histories of dispersal and vicariance that shaped the current distribution of biodiversity.