AI identifies single protein interaction that blocks virus entry into cells before infection

Artificial Intelligence Pinpoints Critical Weakness in Viral Fusion Protein

Scientists at Washington State University have identified a way to block virus entry into cells by targeting a single critical amino acid interaction within a viral fusion protein 1

1

. The research, published in the journal Nanoscale, represents a shift in approach for antiviral therapies, focusing on preventing infection before viruses can slip inside cells where most current drugs begin their work 3

3

.

The team, led by Professor Jin Liu from the School of Mechanical and Materials Engineering and Professor Anthony Nicola from the Department of Veterinary Microbiology and Pathology, used artificial intelligence and molecular simulations to sift through thousands of possible molecular interactions within herpes viruses 2

2

. "Viruses are very smart," Liu explained. "The whole process of invading cells is very complex, and there are a lot of interactions. Not all of the interactions are equally important—most of them may just be background noise, but there are some critical interactions" 1

1

.

Machine Learning Algorithms Accelerate Discovery Process

The researchers developed an algorithm to examine interactions among amino acids, the building blocks of proteins, and then applied machine learning to differentiate and rank which ones were most functionally important 2

2

. This computational approach proved essential because experimentally testing even a single protein interaction can take several months. "It was just a single interaction from thousands of interactions. If we don't do the simulation and instead did this work by trial and error, it could have taken years to find," Liu noted 1

1

.

The study focused on herpes viruses, which rely on a surface fusion protein called glycoprotein B (gB) to merge with and enter cells 3

3

. Scientists have long struggled to understand how this large, complex protein changes shape to make cell entry possible, which helps explain why vaccines for these widespread viruses remain elusive 1

1

.

Laboratory Experiments Confirm AI Predictions Block Virus From Entering Cells

After artificial intelligence identified the key amino acid, the research team moved to laboratory experiments. By introducing a targeted mutation to this specific amino acid, they found that the virus could no longer successfully fuse with cells, effectively blocking the herpes virus from entering cells altogether 1

1

. "The combination of theoretical computational work with the experiments is so efficient and can accelerate the discovery of these important biological interactions," Liu said 2

2

.

Liu emphasized that the value of artificial intelligence in the project was not that it uncovered something unknowable to human researchers, but that it made the search far more efficient. "In biological experiments, you usually start with a hypothesis. You think this region may be important, but in that region there are hundreds of interactions," he told Decrypt. "You test one, maybe it's not important, then another. That takes a lot of time and a lot of money" 3

3

.

Implications for Accelerating Biological Discoveries Beyond Antiviral Therapies

The research, which began more than two years ago shortly after the COVID-19 pandemic and was funded by the National Institutes of Health, points to broader applications beyond virology 3

3

. Liu indicated that the same computational framework could be applied to diseases driven by altered protein interactions, including neurodegenerative disorders such as Alzheimer's disease. "The most important thing is knowing which interaction to target," Liu explained. "Once we can provide that target, people can look at ways to weaken it, strengthen it, or block it" 3

3

.

While the team confirmed the importance of this specific interaction, many questions remain about how the mutation changes the structure of the full viral fusion protein. The researchers plan to continue using simulations and machine learning algorithms to better understand how small molecular changes ripple through entire proteins. "There is a gap between what the experimentalists see and what we can see in the simulation," said Liu. "The next step is how this small interaction affects the structural change at larger scales" 1

1

. The project was conducted by Liu, Professors Prashanta Dutta and Anthony Nicola, along with PhD students Ryan Odstrcil, Albina Makio, and McKenna Hull 1

1

.