New AI model designs proteins to deliver gene therapy

Researchers at the University of Toronto used an artificial intelligence framework to redesign a critical protein involved in the delivery of gene therapy.

The study, published in Intelligence of natural machinesdescribes new work optimizing proteins to attenuate immune responses, thereby improving the effectiveness of gene therapy and reducing side effects.

“Gene therapy is extremely promising, but the body’s pre-existing immune response to viral vectors greatly hinders its success. Our research focuses on hexons, a fundamental protein of adenoviral vectors which, aside from the immune problem, holds enormous potential genetic mutation therapy,” explains Michael Garton, assistant professor in the Institute of Biomedical Engineering in the Faculty of Applied Science and Engineering.

“Immune responses triggered by serotype-specific antibodies pose a significant obstacle to directing these vehicles to the correct target; this can lead to reduced effectiveness and serious adverse effects.”

To solve this problem, Garton’s lab used AI to design custom variants of hexons distinct from natural sequences.

“We want to design something that is distant from all human variants and, by extension, unrecognizable by the immune system,” explains the doctor. candidate Suyue Lyu, who is the lead author of the study.

Traditional methods of designing new proteins often involve extensive trial and error as well as increasing costs. By using an AI-based approach to protein design, researchers can achieve a higher degree of variation, reduce costs, and quickly generate simulation scenarios before focusing on a specific subset of targets for experimental tests.

Although many protein design frameworks exist, it can be difficult for researchers to properly design new variants due to the lack of available natural sequences and the relatively large size of hexons, consisting of an average of 983 amino acids. .

With this in mind, Lyu and Garton developed a different AI framework. Named ProteinVAE, the model can be trained to learn the characteristics of a long protein using limited data. Despite its compact design, ProteinVAE exhibits generative capacity comparable to larger available models.

“Our model leverages pre-trained protein language models for efficient learning on small datasets. We also incorporated many custom engineering approaches to make the model suitable for generating long proteins,” explains Lyu, adding that ProteinVAE was intentionally designed to be lightweight.

Lyu adds: “Unlike other considerably larger models that require high computing resources to design a long protein, ProteinVAE supports fast training and inference on any standard GPU. This feature could make the model more user-friendly for other academic laboratories. Our AI model, validated through molecular simulation, demonstrates the ability to modify a significant percentage of the protein surface, thereby avoiding immune responses.

The next step is experimental testing in a wet lab, says Lyu.

Garton believes the AI ​​model can be used beyond gene therapy protein design and could likely be expanded to support protein design in other disease cases as well.

“This work indicates that we are potentially capable of designing new subspecies and even species of biological entities using generative AI,” he says, “and these entities have therapeutic value that can be used in other new medical treatments.