Abstract 3761: AI-GPT-driven design of novel lipid nanoparticles for targeted and safe mRNA-based cancer immunotherapy

Abstract The development of mRNA vaccines for cancer immunotherapy is limited by challenges in achieving tumor-specific delivery, enhancing safety, and maximizing immune responses. Traditional lipid nanoparticles (LNPs) lack the precision required for targeted delivery, often resulting in off-target effects and systemic toxicity. To address these challenges, we developed an AI-powered lipid discovery pipeline that integrates advanced machine learning (ML) algorithms with Generative Pre-trained Transformer (GPT) models to design and optimize novel ionizable lipids specifically for cancer vaccine delivery. The pipeline leverages GPT-driven molecular design to generate libraries of diverse lipid candidates by analyzing patterns in lipid chemistry datasets and identifying novel molecular architectures tailored to mRNA delivery. Predictive ML models evaluate these candidates for key properties, including ionization efficiency, immune activation potential, and toxicity. By combining GPT’s generative capabilities with ML’s predictive accuracy, the pipeline accelerates the identification of high-performing lipids while reducing discovery timelines by 80%. Using this approach, two novel lipids, RNAi-396 and RNAi-PMA, were identified and validated. RNAi-396 demonstrated lymph node-specific accumulation, essential for localized immune activation, while RNAi-PMA preferentially targeted the spleen, significantly reducing injection site inflammation compared to SM-102, a commonly used clinical lipid. Using RSV mRNA as a model payload, RNAi-396 enhanced IgG titers by twofold, and RNAi-PMA achieved a remarkable 10.6-fold increase over SM-102. ELISpot assays confirmed RNAi-396’s ability to elicit robust T cell responses comparable to SM-102, while RNAi-PMA boosted IFNγ+ splenocyte counts by fivefold, reflecting enhanced cytotoxic T cell activation. The GPT-driven pipeline not only generated novel lipid motifs but also identified structure-property relationships to optimize encapsulation efficiency, biodistribution, and immune modulation. Acute toxicity studies revealed no adverse effects for RNAi-396 or RNAi-PMA, underscoring their clinical safety. The AI-GPT-enabled platform seamlessly iterated molecular designs, integrating feedback from computational and experimental evaluations to refine lipid candidates. This study demonstrates the transformative potential of an AI-GPT-driven pipeline for mRNA-LNP development, combining precision molecular design with rapid optimization to advance cancer immunotherapy. The superior performance and safety profiles of RNAi-396 and RNAi-PMA establish a robust foundation for the next generation of targeted, safe, and effective mRNA-based treatments, marking a significant milestone in the application of AI in precision medicine. Citation Format: Evan Lewoczko, Zachary Dorsey, Yiqing Zou, Renxiang Chen, Yong-Sik Bong, Ezra Chung, David Brown, Steven Long, Dong Shen. AI-GPT-driven design of novel lipid nanoparticles for targeted and safe mRNA-based cancer immunotherapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3761.