Advances in Machine Learning-Enhanced PBPK Models for Brain-Targeted Drug Delivery via Nanocarriers: A Comprehensive Review

Nanostructured drug-delivery materials—liposomes, polymeric nanoparticles, dendrimers, and inorganic carriers—have become central to pharmaceutical strategies for crossing the blood–brain barrier (BBB), where most candidate therapeutics fail to reach their targets. Their biological performance hinges on a coupled chain of vascular transport, BBB translocation, tissue diffusion, cellular uptake, and intracellular release, each of which is shaped by the nanocarrier’s size, surface chemistry, charge, and ligand functionalization. Physiologically based pharmacokinetic (PBPK) models describe this chain mechanistically but are limited by parameter uncertainty, simplified representations of the BBB, and coarse regional resolution. Machine learning (ML) can close these gaps by extracting nonlinear structure–transport–exposure relationships from heterogeneous experimental and clinical datasets. This review examines emerging ML–PBPK hybrid frameworks for predicting the brain biodistribution of nanostructured drug carriers. We compare regression, kernel, and deep learning approaches for parameter inference, model correction, and surrogate modeling; assess strategies for feature selection, uncertainty quantification, and interpretability; and discuss documented failure cases that bound the conditions under which these methods can be trusted. The review closes with recommendations on dataset standardization, software platform selection, and the responsible use of generative AI in pharmaceutical modeling, thus providing guidance for translating nanostructured material design into safer, more effective brain-targeted therapies.