When a virus or parasite replicates, it accumulates mutations. Over time, these genetic shifts create a vast, complex landscape of variants. To track an outbreak, design an effective vaccine, or understand why a specific strain causes severe disease, we cannot treat these variants as isolated sequences. We must reconstruct their shared history.
Phylogenetic tree reconstruction allows researchers to map these mutations back to a common ancestry, organizing viral diversity into distinct genotypes, subtypes, and lineages. By understanding the mathematical and biological intuition behind these trees, we can see how genomic data transforms into an actionable epidemiological roadmap.
The Core Challenge: From Sequences to Lineages
Imagine sequencing a pathogen isolated from a patient. On its own, a string of nucleotides tells us very little. To make it useful, we must perform comparative lineage analysis—aligning the new sequence against a reference set of established genotypes to determine its evolutionary relationships.
This classification is crucial because genetic lineages often dictate clinical outcomes. For instance, in Complete Nucleotide Sequence of Hepatitis B Virus Isolated from Two Infants with Fulminant Hepatitis, Kenji Abe et al. used lineage analysis to identify genotype-specific pathogenic mechanisms in vulnerable infants. Similarly, Boris Renjifo et al. relied on phylogenetic analysis of the HIV-1 env gene in In-utero transmission of quasispecies among human immunodeficiency virus type 1 genotypes to categorize viral sequences into distinct subtypes and study mother-to-infant transmission.
To systematically categorize this diversity, researchers rely on two primary algorithmic paradigms: distance-based clustering and character-based statistical inference.
Distance-Based Clustering: Neighbor-Joining
The most straightforward way to build a tree is to calculate the "genetic distance" between every pair of sequences—essentially counting how many nucleotides differ.
The Neighbor-Joining (NJ) algorithm takes this distance matrix and progressively pairs the closest sequences together, building a tree from the bottom up. Because it relies on a pre-calculated matrix rather than evaluating every individual nucleotide position simultaneously, it is computationally highly efficient.
This speed makes NJ ideal for establishing large-scale public databases and rapid subtyping. For example, in Establishment of public norovirus genome database in Hong Kong., Wai K. Leung et al. utilized neighbor-joining to classify human norovirus strains into specific genotypes for a centralized public repository. Similarly, Yang Jiao et al. applied neighbor-joining trees in HIV-1 transmitted drug resistance-associated mutations and mutation co-variation in HIV-1 treatment-naïve MSM from 2011 to 2013 in Beijing, China to map the distribution and evolutionary history of drug-resistant strains within specific demographic groups.
However, neighbor-joining has a major drawback: it collapses all sequence data into a single distance metric, throwing away the specific context of which mutations occurred where.
Statistical Rigor: Maximum Likelihood
To overcome the limitations of distance matrices, researchers turn to character-based methods, most notably Maximum Likelihood (ML). Instead of compressing the data, ML evaluates the specific nucleotide or amino acid at every single position across the alignment.
Why Maximum Likelihood Works
The core intuition behind Maximum Likelihood is generative: Given a specific tree structure and an explicit model of genetic mutation, what is the probability that this tree would evolve into the exact sequences we observe today?
Instead of asking which tree is the shortest or simplest, ML calculates the likelihood function for thousands of potential tree topologies and branch lengths. It incorporates a substitution model that accounts for biological realities—such as the fact that transitions (e.g., A to G) occur more frequently than transversions (e.g., A to T). The algorithm then selects the tree architecture that maximizes this probability, making our real-world data the least surprising outcome.
This statistical robustness is why ML is the gold standard for resolving complex or messy viral classifications. In Time to Harmonize Dengue Nomenclature and Classification, Lize Cuypers et al. argued that constructing maximum-likelihood trees from whole-genome sequences is essential to provide robust support for classification and clear up global nomenclature confusion.
ML trees excel at capturing subtle evolutionary trajectories across a wide variety of organisms, as demonstrated by several recent studies:
- Long-term adaptation: Ruwini Rupasinghe et al. used ML in Molecular Evolution of Porcine Reproductive and Respiratory Syndrome Virus Field Strains from Two Swine Production Systems in the Midwestern United States from 2001 to 2020 to track two decades of viral adaptation and positive selection.
- Emerging pathogens: The authors of Epidemiology and Evolution of Emerging Porcine Circovirus-like Viruses in Pigs with Hemorrhagic Dysentery and Diarrhea Symptoms in Central China from 2018 to 2021 used ML to visualize genetic divergence and common ancestry in emerging circovirus-like strains.
- Novel wildlife reservoirs: Hayato Harima et al. employed ML in An African tick flavivirus forming an independent clade exhibits unique exoribonuclease-resistant RNA structures in the genomic 3′-untranslated region to place a newly discovered tick-borne virus within the broader flavivirus family tree.
- Eukaryotic parasites: Beyond viruses, Wei Zhao et al. applied ML in Molecular prevalence and subtype distribution of Blastocystis spp. among children who have diarrheia or are asymptomatic in Wenzhou, Zhejiang Province, China to reconstruct relationships among parasite sequences and identify complex transmission routes.
Applications in Global Surveillance
Reconstructing these lineages allows scientists to solve three critical problems in public health and biosecurity.
1. Identifying Unknown Viruses
When a completely novel virus is discovered, researchers cannot match it to a known genome. Instead, they sequence highly conserved genes—like the RNA-dependent RNA polymerase (RdRp), which acts as an evolutionary fingerprint.
In Virome characterization of field-collected rodents in suburban city, Zhenyu Hu et al. leveraged the structural consistency of replication machinery to systematically categorize dozens of previously unknown rodent viruses into established viral families.
2. Assessing Zoonotic Risk
Phylogenetic trees can visually expose when a pathogen has jumped from animals to humans. By building trees that contain both human and animal viral isolates, researchers look for "monophyletic clustering"—meaning the animal and human strains share an exclusive common ancestor.
In Non-Human Primates Harbor Diverse Mammalian and Avian Astroviruses Including Those Associated with Human Infections, Erik A. Karlsson et al. used this comparative sequence genotyping to map unknown primate isolates against a reference library, proving that non-human primates harbor astroviruses capable of crossing species barriers.
3. Uncovering Recombination and Vaccine Evasion
Viruses do not just evolve through slow, linear mutations; they also undergo recombination, where two different strains infect the same host and swap large chunks of their genomes.
When recombination happens, different genes within the same virus will have completely different evolutionary histories. By constructing separate phylogenetic trees for different genetic regions, scientists can spot these anomalies.
For instance, in Vaccine‐Associated Recombination and Virulence Heterogeneity of GI‐19 (QX Type) Infectious Bronchitis Virus in Eastern and Southern China, 2024−2025, Zijian Dai et al. built trees comparing wild isolates against vaccine strains, revealing that viral evolution via recombination was actively outpacing current poultry vaccines. Similarly, Long Zhou et al. combined neighbor-joining and maximum composite likelihood methods in Identification of a Novel Recombinant Type 2 Porcine Reproductive and Respiratory Syndrome Virus in China to isolate a highly mosaic recombinant strain, demonstrating how rapid genetic swapping shapes the trajectory of agricultural epidemics.
Key Foundational Work
Complete Nucleotide Sequence of Hepatitis B Virus Isolated from Two Infants with Fulminant Hepatitis (2003) — contributed 1 concept elements to this direction.
In-utero transmission of quasispecies among human immunodeficiency virus type 1 genotypes (2003) — contributed 1 concept elements to this direction.
Establishment of public norovirus genome database in Hong Kong. (2010) — contributed 1 concept elements to this direction.