Speaker
Description
Microfluidic spheroid-on-chip platforms are widely used for high-throughput drug screening, providing controlled microenvironments for studying tumor–drug interactions. However, drug delivery to tumor spheroids in these devices is governed by complex fluid flow and mass transport processes, making device design challenging. Consequently, their design has often relied on time- and resource-intensive experimental trial-and-error approaches. Here, we develop a computational framework for modeling drug transport in multi-well spheroid-on-chip systems. A computational fluid dynamics model is constructed to simulate fluid flow and drug transport within a microfluidic array, capturing key mechanisms including advection, diffusion, and binding. The model is validated against published experimental data. Simulation results reveal how device geometry, fluid velocity, and tumor- and drug-related parameters influence drug distribution and local shear stress within the wells. Importantly, the simulations predict device-induced heterogeneity in drug exposure and shear stress across spheroids in different wells. To address the need for well-to-well consistency in drug screening platforms, simulation data from parametric sweeps are used to train a neural network surrogate model. The surrogate model is then integrated into a Bayesian optimization framework to identify device configurations that minimize variability in drug exposure across wells, which is essential for reliable drug screening.
