Artificial intelligence can help scientists solve complex biological puzzles, but even advanced systems have limits when the available information is incomplete. A recent study of aphid proteins illustrates how combining computational predictions with genetic evidence can overcome a difficult research problem.

Scientists found that thousands of rapidly changing proteins share a common structural foundation, offering new clues about how these tiny insects influence plant development.

Uncovering a Hidden Protein Family

Aphids can influence the development of certain plants by introducing specialized proteins into their tissues. In some cases, this process causes plants to form structures called galls, which provide a place for the insects to live and feed their offspring.

Scientists have been studying a group of these molecules known as BICYCLE proteins. Their name refers to a repeating pattern involving the amino acid cysteine. Although researchers had identified the proteins, understanding their roles remained difficult because their genetic sequences differed substantially from familiar proteins recorded in scientific databases.

Normally, researchers compare an unfamiliar protein's amino acid sequence with previously studied examples. Similarities can offer clues about its shape and biological function. However, this approach was less effective for BICYCLE proteins because they had changed so extensively over evolutionary time.

Why the AI Prediction Failed

To investigate, the research team worked with structural biologists to determine the physical arrangements of two BICYCLE proteins through X-ray crystallography. This laboratory technique helps scientists establish how atoms are organized within a molecule. The resulting structures provided an important reference for testing AlphaFold2, an AI system designed to predict protein shapes from their amino acid sequences.

Initially, AlphaFold2 failed to reproduce the experimentally determined structures. The problem was not simply a lack of computing power. The system also relies on evolutionary information, including patterns shared by related proteins, to improve its predictions. Because BICYCLE proteins were so unusual, the available databases did not provide enough useful comparisons. The AI therefore lacked an important source of biological context.

Evolution Supplies the Missing Clues

The researchers addressed the problem by collecting additional genetic information from different aphid species. Their work included field sampling in the United States and Japan, followed by genome sequencing to reveal relationships among the proteins.

These comparisons supplied evolutionary evidence that had been missing from the original analysis. With the additional information available, AlphaFold2 successfully predicted a structure matching the one established through crystallography. The team then extended the approach across seven aphid species, generating approximately 2,400 high-confidence structural predictions.

A consistent feature emerged from the results: the proteins shared a structural arrangement known as a saposin-like fold. In simple terms, they appeared to follow a common three-dimensional blueprint even though their individual sequences had changed considerably. The finding revealed a connection that conventional sequence comparisons had struggled to identify.

One Structure, Many Possible Functions

Sharing a structural foundation does not necessarily mean that proteins perform identical tasks. The researchers found substantial variation in the chemical properties of the proteins' outer surfaces. Some regions differed in their electrical characteristics and interactions with water. Attempts to divide the proteins into clearly defined groups based on these properties were unsuccessful.

This diversity suggests that BICYCLE proteins may interact with plant molecules in many different ways. Rather than relying on one specific mechanism, the proteins could influence several biological processes inside plant cells. The researchers also proposed that rapid protein evolution may help aphids maintain these interactions while making their molecules harder for plants to recognize. However, the precise functions of individual proteins still require further investigation.

What the Discovery Means for Future Research

The study, published in the Proceedings of the National Academy of Sciences in September 2026, offers a useful approach for examining proteins that have changed too much for conventional comparison methods to recognize easily. Its importance extends beyond aphids.

Other biological systems also contain rapidly evolving proteins that remain difficult to characterize. Combining laboratory measurements, genetic sequencing, evolutionary analysis, and AI predictions may help researchers investigate these molecules more effectively. Nevertheless, predicted structures are not a substitute for experimental evidence.

The investigation shows that AI can become more effective when researchers provide the right biological context. By combining experimentally determined structures with genetic information from multiple aphid species, the team uncovered a shared blueprint hidden beneath extensive protein variation.

The next challenge is to determine how these molecules influence plant cells and whether their diverse structures correspond to different biological roles. As research continues, the study may help scientists better understand complex plant interactions while demonstrating how evolutionary knowledge can strengthen AI-assisted discoveries.