Objectives
In vitro lipid nanoparticle (LNP) cell-association assays are widely used for formulation screening, but measured signal depends on initial dose, assay geometry, particle transport, and saturation at the cell layer. As a result, parameters estimated from a single exposure condition may not remain predictive when the dose changes. We used a model-informed workflow to evaluate how multi-dose experimental design affects parameter identifiability and prediction of LNP-cell association under untested exposure conditions.
Methods
Time-course LNP-cell association data were generated in HepG2 monolayers using flow cytometry across three initial dose conditions: 0.5x, 1x, and 2x. A transport-association PDE-ODE model was generated to describe sedimentation-diffusion in the culture well coupled to saturable association at the cell surface. Two association kinetic hypotheses were compared: a base capacity-limited model and an uptake-limiting extension in which effective association decreases at higher local particle concentration. Models were evaluated using shared multi-dose calibration, profile-likelihood identifiability analysis, leave-on-dose-out prediction, and prediction of an independent dose dataset without recalibration.
Results
Single-dose calibration produced dose-dependent variation in inferred association parameters, suggesting that individual exposure conditions. However, multi-dose calibration improved parameter consistency and provided a clearer basis for comparing the two kinetic hypotheses. Profile-likelihood analysis showed that the additional parameter term in the uptake-limiting model was weakly constrained when dose information was limited but became better supported when multiple dose conditions were included. In leave-one-dose out prediction, both models reproduced the general time-course trends, while the uptake-limiting model improved prediction at dose-range boundaries. We further validated the calibrated parameters against an independent dose experiment, supporting parameter transferability.
Conclusion
This workflow shows that multi-dose, model-informed analysis can improve identifiability and prediction of in vitro LNP-cell association kinetics. By separating transport, exposure, and association effects, the approach supports more reliable cross-condition prediction and provides a pharmacometrically relevant framework for nanoparticle screening.