New AI tool predicts structures for hundreds of viruses

Bridging the data gap

Researchers at the University of Minnesota's Hormel Institute in Austin have developed a new tool called FoldaVirus. This software is designed to predictively model the physical structures of various viruses. It aims to provide a much larger library of data for scientists who are fighting infectious diseases.

Currently, there are more than 280,000 known viruses in existence. However, the amount of detailed information we have on their physical shapes is quite limited. As of 2024, only 156 of those viruses have had their structures confirmed through cryoEM bioimaging, which is a high-resolution technique used to see tiny biological structures.

The FoldaVirus tool addresses this gap by modeling nearly 700 viruses so far. By providing these models, the researchers hope to give the scientific community a clearer picture of the viral landscape. This work was recently published in the journal Proceedings of the National Academy of Sciences.

How the AI models viruses

The FoldaVirus tool is built upon the AlphaFold artificial intelligence system, which was originally developed by Google. AlphaFold is a well-known AI system used to predict the 3D shapes of proteins. The researchers adapted this technology to help visualize how different viruses are put together.

To create a model, the tool looks at existing information about a specific virus and the broader family of viruses it belongs to. It then uses that data to predict what the structure of the target virus looks like. This allows the system to generate a digital map of a virus even when physical imaging has not been performed.

Every model produced by the tool is given a score. This score indicates how likely the predicted structure is to match the actual physical reality of the virus. By providing these scores, the tool helps scientists understand which models are the most reliable for further study.

From models to medicine

Knowing the exact structure of a virus is a critical step in medical research. If scientists can see the shape, they can identify the specific regions a virus uses to interact with human cells. Identifying these interaction points is the primary way researchers begin to design effective treatments.

Vijay Reddy, an associate professor at the Hormel Institute, explained that this structural information is used to develop various therapeutics. These can include vaccines, neutralizing antibodies, or viral vectors. Viral vectors are tools used to deliver specific genes to cells to prevent the spread of harmful viruses.

The goal is to turn digital models into physical solutions for public health. By understanding the

A global repository for research

The work on FoldaVirus builds upon a broader infrastructure established by the research team. Two years ago, Reddy's lab launched the Virus World Database, which serves as a global repository for information on viruses. This database provides a centralized location for researchers to access and share data about different viral species.

By combining the database with the new AI modeling tool, the institute creates a comprehensive pipeline for virology. Scientists can use the repository to find existing data and then use the AI to generate new structural models. This integrated approach helps the scientific community move faster when identifying new threats.

Speeding up the fight against disease

The primary goal of FoldaVirus is to help virologists work more efficiently. Reddy noted that the tool allows scientists to build virus models seamlessly in just a few hours. This speed is significant because it enables researchers to formulate working hypotheses and generate reagents much faster than traditional methods allow.

The tool is designed to be a practical asset in the fight against viruses and associated diseases. By providing these models quickly, the institute helps researchers focus on the most important parts of the virus. This allows for a more direct path toward creating life-saving interventions for devastating diseases.