Raygun: AI Shrinks Proteins Without Losing Function, Duke Study in Nature Shows

Researchers at Duke University School of Medicine published a generative model called Raygun in Nature on July 29, 2026. It shrinks natural proteins by 10–25% — and in some cases by more than half — while preserving their three-dimensional structure and functional sites. Media outlets have already nicknamed the tool a "molecular shrink ray."

How the "shrink ray" works

Raygun is a template-based model: it takes an existing natural protein as a template and introduces amino acid substitutions and insertions/deletions (indels), rather than designing a molecule from scratch. It builds on protein language models trained on millions of natural sequences. These models convert a chain of amino acids into a standardized mathematical representation, which Raygun then works with.

Control comes down to two knobs. The first sets how strongly to alter the sequence; the second sets the direction in which to move the size — making the protein shorter or longer. Architecturally, Raygun is an autoencoder that compresses a variable-length input into a fixed-size representation.

• Developer — the lab of Rohit Singh, Duke University School of Medicine
• Publication — Nature, July 29, 2026 (DOI 10.1038/s41586-026-10842-8)
• Size reduction — 10–25%, sometimes more than 50%
• Model input — the source sequence, a "noise" parameter, and a target length
• It can also extend proteins beyond their natural size

What the experiments showed

The team tested Raygun on three classes of proteins: fluorescent proteins (used for imaging), cellular enzymes, and variants of epidermal growth factor (EGF). EGF is involved in wound healing and in signaling pathways linked to cancer. For EGF, the team ran receptor-binding experiments and confirmed that the shortened variants retain function.

The key point of these results is that the miniature versions do not lose their working sites. The model introduces considerable diversity into the sequence but preserves functional sites and predicted structural integrity. Unlike models that generate proteins from scratch, Raygun relies on an existing natural template, which raises the chance that the shortened molecule will turn out to be functional.

"There is a language that determines how a protein's amino acid sequence gives rise to its shape and function, but scientists do not yet fully understand that language," —

Rohit Singh, assistant professor of biostatistics, bioinformatics, and cell biology at Duke University.

Why shrink proteins

Compact proteins matter primarily for gene therapy: many therapeutic proteins are too large to fit into viral vectors with limited capacity. Shrinking a molecule without losing its function removes that constraint and widens the range of accessible targets.

Beyond gene therapy, the authors point to the design of protein sensors and other biotech tools. According to Singh, protein language models work like a translator:

"Protein language models act as a kind of translator: they learn patterns from millions of sequences and link those patterns to biological structure and function," —

Rohit Singh, Duke University.

What this means

Raygun turns protein redesign from slow manual engineering into a controllable process: you set a target length and a degree of change, and you get a working variant. This brings "directed evolution" of proteins closer to practical problems in medicine and biotechnology.

Frequently asked questions

What is Raygun?

Raygun is a generative AI model for protein redesign from Duke University, published in Nature on July 29, 2026. It takes a natural protein as a template and creates shortened, lengthened, or heavily modified versions of it while preserving function.

How much does Raygun shrink proteins?

Raygun cuts proteins by 10–25%, and in some cases by more than 50%, while preserving the predicted structure and functional sites. The model can also extend proteins beyond their natural size.

Where is Raygun used?

The main application is gene therapy, where compact proteins are easier to package into viral vectors. The authors also point to the design of protein biosensors and other biotechnology tools.