In the study of evolution, the most difficult relationships to untangle are often those that formed the fastest. When species diverge in rapid succession—a process known as evolutionary radiation—the "branches" on the tree of life become extremely short. These short internal branches represent a fundamental challenge for phylogenetic inference: they contain very little genetic signal, yet they are the keys to understanding major biological transitions.
The Signal Problem in Rapid Diversification
The primary obstacle in resolving rapid radiations is the lack of time for distinct genetic mutations to accumulate. As Constantinos Daskalakis et al. describe in Phylogenies without Branch Bounds: Contracting the Short, Pruning the Deep, these short evolutionary intervals produce signals so weak they become indistinguishable from simultaneous branching events (polytomies).
This ambiguity is not merely a theoretical concern; it appears across diverse biological scales. In the study of human history, J. Víctor Moreno-Mayar et al. found that rapid ancestral splits make it difficult to distinguish between competing models of dispersal, such as the exact geographic origin of Native American founding populations in Terminal Pleistocene Alaskan genome reveals first founding population of Native Americans. Similarly, Xiaoling Fan et al. noted in Clarification of the Phylogenetic Framework of the Tribe Baorini that standard multi-locus datasets often fail to resolve deep history when internal branches are compressed by rapid diversification.
The Inverse Square Scaling Law
To overcome this lack of signal, researchers must increase the volume of genomic data. However, the amount of data required does not grow linearly as branches get shorter. Instead, it follows a specific mathematical relationship known as sample complexity.
Recent theoretical work has established that the number of genes (or loci) required to accurately reconstruct a species tree scales with the inverse square of the shortest branch length ($1/f^2$). Gautam Dasarathy et al. demonstrated in Data Requirement for Phylogenetic Inference from Multiple Loci: A New Distance Method that by using distance-based aggregation methods like METAL, researchers can achieve this $f^{-2}$ scaling. This is a significant improvement over earlier methods that required cubic ($f^{-3}$) or quartic ($f^{-4}$) amounts of data.
This scaling law was further validated for popular summary-based methods like ASTRAL. Shubhanshu Shekhar et al. proved in Species Tree Estimation Using ASTRAL: How Many Genes Are Enough? that the required number of gene trees scales with $1/f^2$ and the logarithm of the number of species ($\log n$). This means that while resolving a branch that is ten times shorter requires a hundred times more data, the total number of species in the tree has a relatively small impact on the data requirement.
The Trade-off Between Loci and Sequence Length
If the goal is to resolve a short branch, should a researcher sequence more genes or sequence longer portions of the existing genes? Elchanan Mossel and Sebastien Roch provide a framework for this decision in Distance-based species tree estimation: information-theoretic trade-off between number of loci and sequence length under the coalescent.
They identified a fundamental trade-off: increasing the length of individual sequences provides a square root improvement in the accuracy of distance estimates. Specifically, the number of genes required is inversely proportional to the square of the branch length and the square root of the sequence length ($1/(f^2 \cdot \sqrt{L})$). This sub-linear scaling means that while longer sequences help, they are not a perfect substitute for increasing the number of independent loci.
When Speciation and Coalescence Overlap
The difficulty of resolving short branches is compounded by a biological phenomenon where the process of species diverging (speciation) and the process of individual genes finding a common ancestor (coalescence) begin to overlap.
In Phylogenetic estimation error can decrease the accuracy of species delimitation, Noah M. Reid explains that when the time since speciation is very short, the branching patterns of the species tree and the within-species genealogy become statistically indistinguishable. This creates a "zone of ambiguity" where it is difficult to determine if a genetic split represents a new species or simply variation within a single population. By using Bayesian frameworks to account for this uncertainty, researchers can avoid making definitive but potentially incorrect classifications in these rapid radiation zones.
Key Foundational Work
Phylogenies without Branch Bounds: Contracting the Short, Pruning the Deep (2011) — contributed 1 concept elements to this direction.
Phylogenies without Branch Bounds: Contracting the Short, Pruning the Deep (2009) — contributed 1 concept elements to this direction.