Artificial Intelligence in the Design and Development of Nanoparticle Drug Delivery Systems: A Systematic Review

Artificial intelligence (AI) has rapidly emerged as a powerful tool for accelerating the design and optimization of nanoparticle-based drug delivery systems (NDDSs). By analyzing complex multidimensional datasets, AI models can predict nanoparticle physicochemical properties, biodistribution, and safety profiles more efficiently than traditional experimental approaches.

This systematic review was conducted following PRISMA 2020 guidelines and was registered in PROSPERO (ID: 1148160). Five electronic databases (PubMed, Scopus, Web of Science, ScienceDirect, and SpringerLink) were searched up to January 2025. Studies applying AI approaches, including machine learning, deep learning, and hybrid modeling to nanoparticle design, optimization, biodistribution prediction, or nanotoxicology, were included. Extracted data included nanoparticle platform, AI methodology, prediction targets, validation strategies, and reported performance metrics. Methodological quality was assessed using a framework integrating TRIPOD-AI and Joanna Briggs Institute (JBI) appraisal domains.

A total of 17 unique studies met the established eligibility criteria and were included in the final qualitative analysis. The applications of machine learning or deep-learning methods and models were demonstrated across the various nanoparticle-based platforms to predict the formulation properties, biodistribution, nano-bio interactions, therapeutic materials discovery, and nanoparticle toxicity. The models augmented by artificial intelligence (AI) technology in the physiologically based pharmacokinetic area manifested promising prediction ability for biodistribution, while the abilities differed among the datasets and the prediction endpoints. Nine studies incorporated in vitro or in vivo biological/experimental validation, whereas only four of the 17 included studies used independent external validation.

AI has significant potential to enhance nanoparticle formulation optimization, biodistribution modeling, and toxicity prediction. Nevertheless, translation into clinical practice remains constrained by small datasets, limited external validation, and inconsistent reporting standards. Future research should prioritize standardized datasets, transparent AI models, and prospective validation to advance AI-enabled nanomedicine toward clinical implementation.