Discordant phylogenies: What to do about it

Research scientist and software engineer Winston Ewert recently discussed how a new tool he helped to develop could potentially solve many of the existing problems with discordant phylogenies. AminoGraph uses a software engineering approach to compare amino acid sequences of proteins to determine genetic relatedness and evolutionary connections between different species.

A recent report by Ewert notes that while the majority of the scientific community readily accepts universal common descent, it has produced numerous failed predictions and remains extremely underdeveloped. An alternate model is proposed here, which is based on explaining similarity using the principles of computer software relationships. Observed similarities in this approach are assigned to function-specific modules that may have additional dependencies. The tool, AminoGraph, presented here for the research community has the potential to revolutionize and solve many of the existing problems with discordant phylogenies.

According to Ewert, phylogeny refers to the evolutionary history and relationships among a group of organisms, illustrating their ancestral relationships and the branching patterns of their evolution over time. Discordant phylogeny refers to a situation in which the evolutionary relationships among organisms provide conflicting evolutionary histories.

"A primary way that evolutionary biologists attempt to determine what evolved from what is by studying the genes found in an organism’s genome," Ewert said. "Many genes are found in a wide variety of organisms, and they look at the similarities and differences in those genes to determine what is most related to what."

Ewert says that the report takes a look at discordant phylogenies and argues that a more sophisticated model than an evolutionary tree is necessary to explain the data.

"Not only do different genes give different evolutionary histories, different parts of the same gene can also give different evolutionary histories," Ewert said.

PAUP (Phylogenetic Analysis Using Parsimony), PhyML, RAxML, MrBayes, BEAST (Bayesian Evolutionary Analysis Sampling Trees), and MEGA (Molecular Evolutionary Genetics Analysis) are examples of common descent inference techniques. AminoGraph, according to Ewert's report, is provided here as a novel alternative that takes an amino acid alignment as input and attempts to create a dependency graph that best fits the data.

Ewert explains that a dependency graph is an idea that comes from software engineering, where pieces of code, referred to as modules, are reused in different software projects. Dependency graphs describe the relationships and dependencies between modules. In his paper, Ewert argues that this model could be used to explain the biological world as a nested hierarchy.

In his paper, Ewert said, "We have extended the dependency graph model to amino acid sequences. In so doing, we have offered an explanation for discordant phylogenies and conflicting phylogenetic signals. We have shown that data that is either randomly generated or produced by a simulated branching process does not exhibit these conflicting signals. However, as shown here with prestin sequences, real genetic data can have such conflicting signals. The new AminoGraph tool provides a way for users to explore the conflicting signals and potential dependency graph influences of amino acid sequences."

According to Ewert, there are established cases of discordant phylogeny where different echolocating species have similar changes to their prestin gene.

"This means that such different creatures as bats and dolphins can get grouped together because of similarities in this gene, whereas overall they are very different," Ewert said. "I focused on it because it was such a well-known example."

Ewert says that to AminoGraph uses Bayesian model selection to determine if a dependency graph is a more appropriate method to study the prestin gene.

"This method works by evaluating the complexity of the model (how complicated a dependency graph or tree is) and how well it explains the data" Ewert said. "Using this technique we see that the dependency graph is a better fit to prestin even if we take into account the complexity of the dependency graph."

AminoGraph has helped emphasize that the pattern of similarities is more complicated than previously understood. Ewert cites an example where the AminoGraph tool was able to uncover two different loci of similarities.

"Some are shared between the microbats, and some are shared between one kind of microbat and the echolocating cateacens," Ewert said.

Ewert says that the most important takeaway for the field of phylogenetics are the similarities and differences in living things are "too complicated to be understood in the simple common descent model."

"Instead, we need a more sophisticated model, such as the dependency graph, in order to understand the similarities and differences in genes," Ewert said.

The AminoGraph tool is available for download at dependencygraph.org.

Winston Ewert., AminoGraph Analysis of the Auditory Protein Prestin From Bats and Whales Reveals a Dependency-Graph Signal That Is Missed by the Standard Convergence Model. Biocomplexity (2023). DOI: 10.5048/BIO-C.2023.1