DIGITAL LIFE

Science's leap forward with AI: the map that reveals the secrets of viruses

For decades, understanding the true structure of a virus required years of laboratory work. Now, artificial intelligence models allow us to reconstruct its three-dimensional shape in just a few hours, accelerating the development of vaccines, antivirals, and precision therapies. This advance promises to transform biomedicine.

Just as an architect needs to know the plans for a building before repairing it, virologists need to visualize the exact shape of a virus to find its weak points. Until recently, this task could take months or even years, using complex protein crystallization and electron microscopy techniques. Today, artificial intelligence is radically changing this process.

An international consortium of scientists from Europe, the United States, and Asia has developed a deep learning model inspired by AlphaFold, created by Google DeepMind in 2021. Unlike previous tools, the new system analyzes not only isolated proteins, but also the interactions between different viral components—capsids, envelopes, and RNA complexes.

With millions of known sequences and structures, the algorithm was trained to predict the complete architecture of a virus in hours based solely on its genetic information. It also identifies critical regions where drugs or antibodies can block its action.

Key innovations of the AI-powered map:

-Massive expansion of the virosphere: Traditional methods for virus discovery are time-intensive and limited in scope. By contrast, LucaProt analyzed genetic information from more than 10,000 datasets, leading to the largest discovery of new virus species in a single study.

-Reveals the hidden "dark matter": The AI tool was designed to detect highly divergent RNA viruses that existing bioinformatics tools often miss, exposing a "viral dark matter" that represents a significant hidden portion of life on Earth.

-Predicts viral evolution: Other AI tools, such as the EVEscape model developed at Harvard and Oxford, use genetic and structural data to predict how viruses like SARS-CoV-2 and influenza will mutate to evade the immune system. This capability can help scientists design "future-proof" vaccines and therapies by anticipating future variants.

-Unlocks viral protein structures: By leveraging AI models like AlphaFold, researchers can predict the protein structures of viruses at scale, which is difficult to do with traditional genome analysis alone. This structural mapping reveals deep, distant evolutionary connections that are often hidden by rapid viral mutations.

From prediction to medical impact...The results are impressive: over 90% coincidence between the structures predicted by AI and those confirmed in the laboratory. This paves the way for rapidly mapping emerging viruses—such as coronaviruses, influenza, or hemorrhagic viruses—and anticipating their vulnerabilities before global expansion.

In addition to guiding more effective vaccines, the technology can guide the development of precision antivirals capable of binding to specific points in the viral structure. It will also be possible to test thousands of pharmacological compounds in digital simulations, reducing the time and cost of biomedical research.

A revolution comparable to the human genome...The scientists who published the study in Science compare the breakthrough to the decoding of the human genome. Now, structural biology is entering a new era: the possibility of exploring the "structural universe" of viruses, understanding how they fold, how they evolve, and which regions remain stable over time.

This approach even allows us to predict future mutations and how they might alter the virus's shape, helping to anticipate resistance to antivirals or immune escape from existing vaccines.

Limitations and upcoming challenges...Despite this leap, the authors emphasize that AI does not replace traditional experiments. Its predictions need to be validated with techniques such as electron microscopy or X-ray diffraction. The challenge now is to expand the model to more complex viruses, such as those with double-stranded DNA, and to unify global databases.

A new map to face microscopic enemies...The researchers describe this achievement as having, for the first time, a "Google Earth of viruses": a three-dimensional and dynamic map of the invisible enemy. With this resource, science will be able to anticipate outbreaks, develop more specific vaccines, and better understand viral evolution. As the authors state: "Knowing the exact shape of a virus is half the battle. The other half is knowing how to use that information to protect human health."

mundophone