Researchers at the University of Toronto’s Faculty of Applied Science and Engineering have successfully utilized an artificial intelligence (AI) framework to redesign a crucial protein involved in the delivery of gene therapy. Led by Dr. Michael Garton, this study focuses on optimizing proteins to mitigate immune responses, therefore improving the efficacy of gene therapy and reducing side effects.
This research was published in the recent issue of Nature Machine Intelligence.
“Gene therapy holds immense promise, but the body’s pre-existing immune response to viral vectors greatly hampers its success. Our research zeroes in on Hexons, a fundamental protein in adenovirus vectors, which ––but for the immune problem–– hold huge potential for gene therapy,” explains Dr. Michael Garton, an assistant professor at the Institute of Biomedical Engineering at the University of Toronto and the corresponding author of this research. “Immune responses triggered by serotype-specific antibodies pose a significant obstacle in getting these vehicles to the right target, this can result in reduced efficacy and severe adverse effects.”
To address this issue, Garton’s team aims to use AI to custom-design variants of Hexons that are distinct from natural sequences.
“We want to design something that is distant from all human variants, by extension, unrecognizable by the immune system.” explains Suyue Lyu, a PhD candidate and the lead author of this research.
Traditional methods of designing new protein would take immense trial and error, and often result in high cost. By using an AI-based approach for protein design, researchers can achieve a higher degree of variation, reduce experimental costs, and quickly generate simulation scenarios before homing in on a specific subset of targets for experimental testing.
Although there are numerous protein-designing frameworks existing in the market, the large size of Hexons (983 amino acids on average) and the lack of natural sequences available poses a challenge to properly design new variants. With this challenge in mind, Lyu and team had to come up with a completely different AI framework.
Dubbed ProteinVAE, this model can be trained to learn the characteristics of a long protein on limited data. Despite its compact design, ProteinVAE exhibits a generative capability comparable to substantially larger models available in the market.
“Our model takes advantage of pre-trained protein language models for efficient learning on small datasets, we also incorporated many tailored engineering approaches to make the model suitable for generating long proteins,” said Lyu. “We intentionally designed ProteinVAE to be lightweight. Unlike other considerably larger models that demand high computational resources to design a long protein, ProteinVAE supports fast training and inference on any standard GPUs. This feature could make the model more friendly for other academic labs.”
“Our AI model, validated through molecular simulation, demonstrates the ability to change a significant percentage of the protein’s surface, potentially evading immune responses,” said Lyu, “Our next step would be to move on to wet lab experimental testing.”
Dr. Garton also believes this AI-model can be utilized beyond gene therapy protein design, and likely expanded to protein design in other disease cases as well.
“This work indicates that we are potentially able to design new subspecies and even species of biological entities using generative AI, and these entities have therapeutic value that can be used in novel medical treatments,” said Dr. Garton.