A predictive computational platform for optimizing the design of bioartificial pancreas devices

Encapsulated islets or stem cell-derived insulin-producing cells in bioartificial pancreas (BAP) devices hold promise for treating type 1 diabetes, but their efficacy is hampered by mass transport limitations, particularly oxygen diffusion. This paper introduces SHARP, a computational platform using the stochastic finite element method to model oxygen transport and its impact on cell survival and insulin secretion in BAP devices, accounting for islet size distribution and random localization. The platform reveals the significant influence of islet size distribution variance on device potency. SHARP is then used to optimize device structures and estimate curative cell doses, leading to a new device-specific islet equivalence conversion table and a surrogate machine learning model (SHARP-ML) for rapid prediction of these coefficients.

Alexander U. Ernst, Long-Hai Wang, Scott C. Worland, Braulio A. Marfil-Garza, Xi Wang, Wanjun Liu, Alan Chiu, Tatsuya Kin, Doug O’Gorman, Scott Steinschneider, Ashim K. Datta, Klearchos K. Papas, A. M. James Shapiro, Minglin Ma