Raygun shortens natural proteins while preserving function in cellular assays

Raygun shortens natural proteins while preserving function in cellular assays

On July 29, Nature published a paper on Raygun, a model that redesigns an existing protein sequence to fit a specified length. In cellular assays, six of the eight selected shortened fluorescent proteins remained fluorescent.

A protein is a chain of amino acids that must fold into a functional shape. Its size often creates practical problems. A large fluorescent tag can interfere with the observation of a small protein, while a gene delivered by a viral vector must fit within the vector's limited capacity. Simply removing part of the chain does not work because neighboring amino acids affect one another's positions, causing the protein to lose its shape.

Raygun treats protein length and sequence as a single design problem. The model receives a natural protein, the required length, and a setting that controls the proportion of substitutions. It then proposes amino acid insertions, deletions, and substitutions so that the new chain retains the predicted structure of the original protein. Previous generative methods were particularly effective at creating proteins from scratch or changing individual amino acids. Here, the researchers redesign an existing functional protein.

The authors tested the method on the green fluorescent protein eGFP and the red fluorescent protein mCherry. Raygun generated 70 thousand sequences for each protein. The researchers selected eight variants, introduced them into HEK293 cells, and observed fluorescence above the negative control in six variants after four days. The functional eGFP contained 199 amino acids instead of 238, while mCherry contained 206 instead of 236. Both were shorter than 96% of the fluorescent proteins in the FPbase database.

Reducing protein size frees space within a genetic construct. The authors also produced active variants of TurboID, a protein used to study the proximity of other proteins within cells, as well as variants of epidermal growth factor EGF that bound more strongly to the EGFR receptor than the original EGF. These sequences can serve as candidates for further testing in a specific laboratory system.