The choice of structural template can create apparent links between predicted protein stability and longevity in tetrapods
This study demonstrates that the choice of structural template in protein stability predictions can artificially generate spurious associations between predicted stability and species longevity in tetrapods, highlighting the need for methodological checks like using outgroup templates to avoid such biases.
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 This is an AI-generated explanation of a preprint that has not been peer-reviewed.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
Across the animal kingdom, the clock of life ticks at wildly different speeds. A mouse might live for a few years, while a tortoise can thrive for over a century. For decades, scientists have searched for the molecular reason behind this disparity, wondering if the very building blocks of life—proteins—are somehow more durable in long-lived creatures. Proteins are the workhorses of the body, folding into intricate three-dimensional shapes to perform their tasks. If these shapes become unstable or unravel, the cell suffers. It seemed logical to assume that animals with long lives would have evolved proteins that are exceptionally stable, resisting the wear and tear of time better than those of short-lived species. To test this, researchers have increasingly turned to powerful computer models that predict how stable a protein is based on its structure. However, a new study suggests that the very tools used to measure this stability might be creating an illusion, leading scientists to see patterns that are actually just artifacts of how the computer models are built.
The researchers set out to investigate whether the choice of a "structural template"—a reference shape used as a scaffold for computer predictions—could accidentally create a false link between protein stability and longevity. In their study, they examined two distinct groups of four-legged animals: fifty-seven species of placental mammals and forty species of reptiles and amphibians. For each animal, they looked at thousands of genes and compared the proteins of that living species to a reconstructed version of what their ancient ancestor likely looked like. They then used a computer program called ThermoMPNN to predict how much each difference between the ancestor and the modern animal would destabilize the protein's shape. Crucially, they ran these predictions using different reference shapes, or backbones, to see if the results changed depending on which template they chose.
When the team first ran the analysis using a human protein structure as the template, a pattern emerged that seemed to support the idea of stable proteins for long lives. In the group of mammals, the primates—which include humans and are known for living longer than other mammals of similar size—appeared to have the most stable proteins. The computer predicted that the changes in their proteins were the least likely to cause the structure to fall apart. This looked like strong evidence that long life is linked to molecular durability. However, the researchers suspected that this result might be a trick of the template. Because the human structure was used as the reference, the computer was essentially measuring how similar other animals were to humans. Since primates are closely related to humans, their proteins naturally looked more stable in this specific context, creating a false association with their long lifespans.
To test this suspicion, the researchers repeated the experiment using a completely different template: the protein structure of a platypus, an egg-laying mammal that is distantly related to the other species in the study. When they switched to this new reference, the statistical link between protein stability and lifespan disappeared (P = 0.701), and the apparent advantage for primates was greatly reduced, though a small residual offset remained. The long-lived primates no longer stood out as having the clearly most stable proteins; their scores shifted to match the rest of the group more closely. The same issue appeared in the reptile and amphibian group. When the team used a human template, the slow-evolving turtles and crocodiles, which are also long-lived, appeared to have the most stable proteins, giving a nominal association. But when they used a fish template, which is equally distant in time from all the reptiles, this pattern did not reproduce. The computer was simply favoring the animals that had changed the least since the common ancestor, regardless of how long they lived.
The study concludes that the choice of a structural template can create a convincing but entirely artificial connection between protein stability and longevity. The researchers did not find a robust, real-world link where long-lived animals have inherently more stable proteins than short-lived ones. Instead, they found that the computer models were sensitive to the specific shape used as a starting point. If the starting shape is too similar to a long-lived group, that group will look artificially stable. While the possibility of a moderate effect remains and cannot be excluded, the clear signal that long life equals stable proteins, as previously suggested by some models, appears to be a mirage created by the tools themselves. The authors propose that future studies must use multiple, unrelated reference shapes and check for these biases to ensure they are measuring real biological differences rather than computer artifacts.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.