The Secret language of Herpes and the AI that Learned to Read it

For years, herpesviruses have played a long game with the human body. They slip in quietly, hide for decades, and then, often without warning, wake up and reactivate. Scientists have known the genes involved. What’s been harder to understand is how viruses decide when to stay silent and when to act.

A new study by Bar-Ilan University researchers Nilabja Roy Chowdhury, Deepanway Ghosal, Vyacheslav Gurevich, and Meir Shamay (published in Nature Communications, 2026) offers a powerful new way to answer that question. They built an AI tool that can scan viral DNA and locate the “switches” that control viral gene activity.

The hidden switches inside viral DNA

Genes are like machines. But machines don’t run unless someone flips the switch.

In DNA, those switches are called enhancers. They don’t make proteins themselves. Instead, they control how strongly other genes turn on, when they activate, and under what conditions. In viruses, enhancers are especially tricky to find because viral DNA is short, dense, and often multitasks.

Until now, mapping viral enhancers required slow, complex lab work, and usually only worked in one cell type at a time.

Teaching AI to read DNA like a language
The BIU research team took a different approach. They treated DNA like language.

Just as sentences follow grammar, DNA follows patterns. The researchers trained a language model, similar in spirit to tools used to understand human text, on only six known enhancer sequences from one herpesvirus.

That’s it. Six examples.

They called the model ENHAvir. Then they asked a bold question: could it recognize enhancer “grammar” across other herpesviruses?

Finding the virus’s control center
The answer was yes.

ENHAvir successfully identified real, working enhancers across many human herpesviruses, including Epstein–Barr virus, herpes simplex virus, and cytomegalovirus. The team tested them in the lab and confirmed that these DNA regions truly function as gene switches.

One discovery stood out. In every herpesvirus they examined, the same region kept lighting up: the terminal repeats, the DNA sequences at the very ends of the viral genome.

These regions turned out to be powerful enhancer hubs, acting like master control panels that help decide whether the virus stays dormant or switches into full replication.

A surprise connection to human DNA

An AI trained only on viral DNA could also recognize enhancer regions in human genes. That suggests viruses and humans may share parts of the same regulatory “language” because viruses evolved to hijack our cellular machinery as efficiently as possible.

Why this matters
This study changes how we look at viruses.

It shows that viral behavior is about finely tuned control systems deciding when those genes speak up. It also shows that AI, when trained on small but high-quality biological data, can uncover patterns that are nearly invisible to traditional methods.

In simple terms:

BIU researchers taught AI to listen more carefully and it heard the whispers that tell viruses when to sleep and when to wake.