AI-driven discovery and engineering of human-derived protein scaffolds for mRNA delivery

Safe and effective gene delivery remains a central challenge for therapeutic applications. While non-viral and viral vectors have enabled substantial progress, their reliance on non-human components often triggers immune responses, limiting their use in chronic treatments. Here, we develop DeepDelivery, an artificial intelligence-driven platform to identify human proteins with scaffold potential for mRNA delivery. Proteome-scale screening nominated 512 candidates, and experimental validation show that 12 of 15 selected human proteins functioned as scaffold components that facilitated particle formation and mRNA delivery within a modular engineering framework. Notably, several tripartite motif (TRIM) family proteins, typically linked to antiviral responses, exhibit strong assembly and delivery activity. Model interpretability and perturbation analyses identify TRIM domain features that influenced prediction scores and guide TRIM25-specific domain engineering to improve delivery. Together, these results establish a framework for mining the human proteome for scaffold components of programmable mRNA delivery systems and for further exploring the design space of human-derived delivery architectures. Protein-based carriers could provide biocompatible options for mRNA delivery. Here, authors develop an AI-guided platform to identify and engineer human-derived protein scaffolds that assemble into particles and support mRNA delivery in cultured cells.