Protein-based nanomedicines for cancer theranostics: From supramolecular self-assembly to AI-driven design and applications

Protein-based nanomedicines represent a paradigm shift in cancer theranostics, capitalizing on superior biocompatibility, molecular recognition, and multifunctional adaptability. This review delineates their evolution from supramolecular self-assembly to artificial intelligence (AI)-driven design, emphasizing their transformative role in cancer chemotherapy, phototherapy, chemodynamic therapy, and immunotherapy. Supramolecular strategies, including metal coordination, electrostatic interactions, host-guest chemistry, hydrogen bonding, π-π stacking, and hydrophobic interactions, not only enable precise control over protein assemblies but also facilitate drug delivery and performance. AI tools like AlphaFold 3 and RFdiffusion have accelerated de novo protein design and dynamic interaction prediction, overcoming limitations in structural prototyping. Despite breakthroughs, challenges persist in the mechanistic insights into assembly dynamics, experimental validation of AI-generated constructs, and scalable clinical translation. Future directions prioritize integrated theranostics platforms, multi-omics-guided precision medicine, and synthetic biology. By synergizing supramolecular chemistry, AI, and nanotechnology, this review envisions protein-based nanomedicines as intelligent, adaptive systems poised to redefine paradigms of cancer theranostics.