In a recent study published in the journal Nature Biotechnology, researchers explored the potential of artificial intelligence (AI) techniques to improve gene therapy delivery systems specifically for lung diseases.
Study: Using artificial intelligence to develop gene therapy for the lungs. Image Credit: Viewfoto studio/Shutterstock.com
The research team focused on optimizing lipid nanoparticles (LNPs) through lipid optimization using neural networks (LiON), a deep learning-based novel approach designed to address persistent challenges in pulmonary gene therapy. The goal was to achieve safe and efficient delivery of therapeutic agents to lung tissues, which is a longstanding barrier to advancing treatments for respiratory diseases.
Gene Therapy and the Challenge of Lung Delivery
Gene therapy, a promising treatment for genetic disorders like cystic fibrosis, works by delivering therapeutic genes to cells to restore their function. However, achieving effective delivery to the lungs has been particularly challenging due to the body’s natural defenses against foreign particles. Traditional delivery methods often fail to balance safety and efficiency, which is one reason why no FDA-approved lung gene therapies currently exist.
Lipid nanoparticles (LNPs), however, have emerged as a powerful solution. Their success in delivering mRNA in COVID-19 vaccines has demonstrated their potential for transporting genetic material across cell membranes while improving stability and bioavailability. For pulmonary applications, though, designing LNPs capable of targeting lung tissues remains complex. Their chemical composition plays a key role in determining delivery efficiency, creating a need for better design strategies.
The LiON Approach: AI for LNP Optimization
To address this, the researchers developed LiON, a deep learning model designed to predict how well LNPs can deliver nucleic acids based on their chemical composition. The team compiled a dataset of over 10,000 LNP formulations, including diverse structures and delivery cargoes, to train the model.
Using a graphical neural network, LiON analyzed the chemical properties of LNPs to identify and rank potential lipid candidates. This AI-driven strategy reduces human bias and enhances the design process by integrating large datasets, leading to more accurate predictions than conventional approaches.
Key Findings: Enhancing Gene Delivery
The study showed that LiON accurately predicted the mRNA delivery potential of LNPs. One key finding was the identification of RJ-A30-T01, a lipid that delivered mRNA to the liver with nine times greater efficiency than the previous best-in-class lipid. This improvement highlighted the potential of deep learning in discovering effective lipid formulations.
The authors also synthesized and tested over 1.6 million LNPs generated in silico to identify candidates suitable for pulmonary gene therapy. Among these, fatty-oligomer 32 (FO-32) and FO-35 emerged as the most promising lipids. Rigorous testing demonstrated their exceptional mRNA delivery capabilities across various delivery routes, including intramuscular, intranasal, and intratracheal methods.
Preclinical testing in ferrets, a well-established model for lung research, confirmed that both FO-32 and FO-35 effectively delivered mRNA to the alveoli and conducting airways. These outcomes highlighted the effectiveness of deep learning in designing LNPs, showing that incorporating diverse datasets can lead to the discovery of novel lipid chemistries.
Implications for Gene Therapy
This study opens up new possibilities for treating lung diseases like cystic fibrosis, where direct delivery of therapeutic genes could significantly improve outcomes. Beyond lung applications, the LiON model could also be applied to optimize LNPs for other areas of gene therapy, such as treatments for cancer and genetic disorders.
The integration of deep learning with biotechnology marks a step forward in overcoming delivery challenges and designing safer, more effective therapies.
Moving Forward
While these findings are promising, the researchers emphasize the need for further testing in more advanced models, such as primates, to confirm the safety and effectiveness of FO-32 and FO-35 before progressing to clinical trials. Generating high-quality in vivo data will also help refine predictive models like LiON, enhancing their accuracy and expanding their utility.
By combining machine learning with biotechnology, this research not only improves the design of LNPs but also sets a solid foundation for the future of gene therapy. As these approaches evolve, they have the potential to deliver better treatment options, improve patient outcomes, and push the boundaries of therapeutic development.
Journal Reference
Using artificial intelligence to develop gene therapy for the lungs. Nat Biotechnol (2024). DOI: 10.1038/s41587-024-02491-x, https://www.nature.com/articles/s41587-024-02491-x
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.