🧬 biology

Unstable Species Trees Reveal Fundamental Limits to Phylogenomic Inference

This study demonstrates that extreme instability in species trees across the genome of sagebrush lizards, driven by widespread introgression and geographic barriers, challenges the feasibility of inferring a single evolutionary history and supports recognizing multiple distinct species within the *Sceloporus graciosus* complex.

Original authors: Christopher Blair, Tomas Flouri, Felipe DE MEDEIROS MAGALHÃES, Lauren Chan, Daniel Mulcahy, Sungsik Kong, Adam Leache

Published 2026-07-21

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Christopher Blair, Tomas Flouri, Felipe DE MEDEIROS MAGALHÃES, Lauren Chan, Daniel Mulcahy, Sungsik Kong, Adam Leache

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine trying to figure out the family tree of a massive, sprawling clan of relatives who all look a bit alike and live in a neighborhood where everyone frequently visits each other's houses. In the world of science, this is the job of phylogenetics: the art of drawing the "Tree of Life" to show how different species are related. For a long time, scientists thought that if they just gathered enough DNA data, the tree would become crystal clear. But recently, they've discovered something weird: the more data they collect, the more the branches of the tree seem to wiggle and argue with each other. This is called phylogenomic conflict. It happens because evolution isn't always a straight line; sometimes, different parts of an organism's DNA tell different stories because of ancient mixing (called introgression) or because family members didn't sort out their genetic traits perfectly before splitting up (called incomplete lineage sorting). Understanding this is crucial because if we get the family tree wrong, we might misunderstand how species evolved, where they came from, and even how to protect them today.

Now, enter the sagebrush lizards of the American West, specifically the Sceloporus graciosus group. These lizards are like the chaotic, fun-loving cousins of the animal kingdom, scattered across the rugged mountains and deserts of western North America. A team of scientists decided to take a deep dive into their DNA to finally settle the debate: just how many distinct species are hiding in this group, and what does their family tree actually look like? They didn't just look at a few genes; they used a massive amount of genomic data—over 100,000 tiny snippets of DNA from 83 different lizards—to try and reconstruct their history.

What they found was a bit of a shocker. Instead of finding one single, stable family tree, they discovered that the tree was incredibly unstable. It was as if they were trying to assemble a puzzle, but every time they picked up a different handful of pieces, the picture changed completely. Some chunks of DNA suggested the lizards were related one way, while other chunks insisted on a totally different arrangement. The researchers realized that no single "best" tree could explain the history of these lizards. The conflict wasn't just a mistake in their math; it was a real biological signal. The data suggested that these lizards had a history of rapid diversification mixed with frequent "genetic borrowing" (introgression) between neighboring groups. Essentially, the lizards were so busy swapping genetic material across the landscape that their family history became a tangled web rather than a neat, branching tree.

The study explicitly argues against the idea that we can simply force all this data into one single, perfect tree using standard computer methods. In fact, the authors found that the most popular method of just gluing all the DNA together (called "concatenation") actually gave them a misleading picture that didn't fit the reality of the data. They also ruled out the idea that this confusion was just due to bad data or a lack of information; they had plenty of data, and the confusion persisted even when they used the most advanced statistical tools available. The paper suggests, rather than proves with absolute certainty, that the best way to understand these lizards is to accept that their history is a complex network of connections, not a simple line.

So, what does this mean for the lizards themselves? The researchers used this messy, conflicting data to propose a new way of looking at the group. They suggest that what we used to think of as one or two species with a few subspecies is actually a collection of at least six distinct species. Some of these are clearly separated by huge geographic barriers like mountain ranges and deserts, while others, like the highly specialized S. arenicolus (the sand-dwelling lizard), have carved out their own unique path. The study highlights that even with thousands of genetic clues, nature can be so complex that it resists being forced into a single, simple diagram. Instead of a rigid tree, the history of these lizards is more like a braided stream, where different genetic threads weave together and separate over time. This discovery serves as a wake-up call for scientists everywhere: when dealing with rapidly evolving groups, we need to stop looking for one perfect answer and start embracing the messy, beautiful complexity of nature's true history.

Technical Summary: Unstable Species Trees Reveal Fundamental Limits to Phylogenomic Inference

Problem Statement
Phylogenomic conflict—the discordance between gene trees and species trees—is increasingly recognized as a pervasive challenge in evolutionary biology, particularly during rapid radiations. While previous studies have focused on deep evolutionary questions or specific causes like incomplete lineage sorting (ILS) and introgression, there is a lack of comprehensive understanding regarding the stability of species tree inference when using full likelihood-based methods on genome-wide data from closely related, geographically widespread taxa. Specifically, it remains unclear how subsampling strategies, the presence of gene flow, and the limitations of current computational models affect the recovery of a single, robust species topology. This study addresses these gaps by investigating the Sceloporus graciosus group (sagebrush lizards) in western North America, a system characterized by rapid divergence, complex biogeography, and potential cryptic diversity.

Methodology
The authors employed a multi-faceted analytical approach using genomic data from 83 individuals of the S. graciosus complex and two outgroups.

Data Generation: Genomic data were generated via Genotyping-by-Sequencing (GBS), yielding over 100,000 loci (>7 million bp). Mitochondrial DNA (ND4 and concatenated mtDNA) was also sequenced for 100 individuals.
Population Genomics: Population structure was assessed using ADMIXTURE across various $K$ values, with sensitivity analyses performed on missing data thresholds. Isolation-by-distance (IBD) was tested using Maximum Likelihood Population Effects (MLPE) models.
Phylogenomic Inference:
- Concatenation: Maximum Likelihood (ML) trees were inferred using IQ-TREE.
- Coalescent Methods: Species trees were estimated using CASTER, SVDquartets, and BPP (Bayesian Phylogenetics and Phylogeography).
- Genome-Wide Stability Analysis: To assess topological stability, the authors performed extensive subsampling. They partitioned the genome into bins of 1,000 loci (both random subsets and reference-guided ordered bins mapped to a S. arenicolus draft genome). Each bin was analyzed independently using BPP under the Multispecies Coalescent (MSC) model.
- Model Selection for Introgression: The authors utilized BPP's MSC-with-introgression (MSC-I) and MSC-with-migration (MSC-M) models. They constructed network models (N1–N8) with varying numbers of reticulation events and compared their fit using marginal likelihood estimates and Bayes factors.
- Hybrid Detection: Historical hybridization was detected using HyDe, and phylogenetic networks were inferred using SNaQ.
- Species Delimitation: Species boundaries were evaluated using mPTP on mtDNA, and guided delimitation (A10) and genealogical divergence index (gdi) analyses (hhsd) on genomic bins within BPP.
- Spatial Analysis: Effective migration surfaces were modeled using EEMS to visualize barriers to gene flow.

Key Results

Widespread Topological Instability: The study found extreme instability in species tree topologies. Random and reference-guided subsampling of loci failed to consistently recover a single topology. Across genome-wide bins, 33 unique topologies were sampled, with seven distinct topologies accounting for >99% of the posterior mass. No single topology was stable across the entire genome.
Role of Introgression: The instability was attributed largely to rampant introgression. Model selection via Bayes factors strongly favored the most parameter-rich network model (N8), which included multiple introgression events, over strictly bifurcating trees.
Gene Flow vs. ILS: While ILS is present (indicated by short internal branches), the authors argue that a small number of loci affected by gene flow can overwhelm the phylogenetic signal of the remaining loci, distorting species tree inference even in full likelihood frameworks like BPP.
Population Structure and Barriers: ADMIXTURE and EEMS analyses revealed five to seven distinct genetic clusters corresponding to geography. Significant barriers to migration were identified, including the Wasatch Mountains, the Sierra Nevada, the Transverse Ranges, and a major barrier separating S. arenicolus from the rest of the complex in New Mexico/Texas.
Species Delimitation:
- S. arenicolus was consistently recovered as a distinct, monophyletic lineage with no gene flow, supporting its status as a separate species.
- Genomic data supported the recognition of S. g. vandenburgianus (Southern California) and S. g. gracilis (Northern California) as distinct species, based on clustering, lack of significant gene flow, and morphological distinctness.
- The remaining lineages (Colorado Plateau, Central California A/B) showed complex relationships and were recommended to remain as S. graciosus pending further analysis.
Methodological Performance: Concatenation (CASTER) performed poorly, often yielding poorly supported or conflicting topologies. BPP and SVDquartets showed more variability but were generally preferred for rapid radiations, though they too struggled to converge on a single stable tree in the presence of high gene flow.

Significance and Claims
The paper claims to provide the most comprehensive assessment of phylogenomic incongruence at the phylogeographic level to date. Its primary contribution is demonstrating that species tree inference can be fundamentally unstable even when using powerful, full likelihood-based methods that account for ILS, provided that introgression is widespread.

The authors argue that:

Subsampling is Critical: Researchers must perform subsampling experiments to assess the consistency of species tree topologies across the genome, as different subsets of loci can yield statistically supported but conflicting histories.
Uncertainty in Downstream Inference: The practice of estimating a single species tree and using it as a fixed framework for downstream analyses (e.g., divergence dating, biogeography) is problematic in systems with extensive ILS and introgression. The paper advocates for explicitly incorporating species tree uncertainty into evolutionary inferences.
Taxonomic Implications: The study supports a revised taxonomy for the S. graciosus complex, recognizing up to six distinct species, including the elevation of two subspecies to full species status.
Limits of Current Models: The results highlight that current MSC models, even when extended for introgression, face challenges in disentangling complex histories where gene flow is pervasive, suggesting that a "single tree" may not be the best representation of evolutionary history for such groups.

The authors conclude that while phylogenomic data has revolutionized systematics, the field must grapple with the reality that for rapidly radiating, geographically widespread taxa, the "true" species tree may be a reticulate network rather than a bifurcating tree, and that standard inference workflows may need significant adaptation to handle this uncertainty.

Technical Summary: Unstable Species Trees Reveal Fundamental Limits to Phylogenomic Inference

More like this