Kad biologija nadahnjuje matematiku: Skrivene simetrije objašnjavaju zašto široko korištene evolucijske metode mogu dati pogrešne odgovore

Scientists have identified a hidden mathematical symmetry within commonly used evolutionary models that explains why these tools can occasionally yield misleading results. A recent study reveals that certain assumptions embedded in these models allow for multiple distinct evolutionary histories to produce identical observational outcomes, leading to potentially flawed conclusions about biodiversity patterns. The discovery stems from a collaboration between researchers at the Finnish Museum of Natural History and Virginia Tech. Their investigation began with an unexpected observation related to beetle anatomy. While examining the classification of Helictopleurus sicardi, a type of beetle, they encountered a challenge similar to grouping objects based on shared characteristics. This led them to explore a mathematical principle called lumpability, which determines when different states in a Markov model can be grouped without altering the model’s behavior. Building upon this idea, the researchers developed a novel method for representing Markov models, which are foundational to many areas of science. They demonstrated that each discrete-state Markov model can be transformed into an equivalent hidden-state model through a process they term Hidden Expansion. Though the reformulated model appears more complex, it is constructed from simpler, repeating elements. This approach revealed previously undetected mathematical symmetries, transforming a seemingly unsolvable issue into one that could be addressed systematically. Published in Nature Communications, the study highlights a critical flaw in the models long used by evolutionary biologists. These models, which attempt to predict how traits influence species diversity, often produce conflicting results due to inherent ambiguities. The researchers found that the same hidden symmetries affecting basic classification tasks also impact advanced models used to study biodiversity. This means that the occasional errors observed in these models are not random statistical anomalies but rather a consequence of deeper structural issues. “This mathematical property had gone unnoticed despite decades of research on Markov models,” notes Sergei Tarasov, one of the lead authors. He explains that the realization came after the team applied their findings to real-world datasets. One such case involved a published study on stick insects, which examined whether the evolution of male weaponry influenced speciation rates. The original analysis concluded there was no evidence linking weapon development to increased diversification. However, when the researchers reanalyzed the data using a wider array of models, standard statistical techniques suggested otherwise. This discrepancy underscores the potential pitfalls of relying solely on conventional methods. “For years, researchers could see the symptoms,” Tarasov adds. “Our work uncovered the underlying mathematical cause. Once we recognized that many fundamentally different evolutionary histories can produce exactly the same observations, it became clear why these methods can sometimes point to a convincing but ultimately incorrect biological explanation.” The implications of this finding extend beyond evolutionary biology. By exposing a general limitation in the way these models operate, the study offers insights that could improve the accuracy of predictive tools in fields ranging from genetics to ecology. As the researchers continue refining their approach, they hope to develop more robust frameworks that account for these hidden symmetries, ensuring that future analyses better reflect the true complexity of life's evolutionary journey.