AI-driven ANN and RSM-CCD integrated optimization of cinnarizine-domperidone bilayer tablet: In-vitro evaluation and in-silico PBPK modeling using GastroPlus®

Background: Response surface methodology coupled with the design of experiments identifies the optimal response surface function related to selected independent factors. Due to their rigid structures, they lack the learning capability from the developed response function. In contrast, a more advanced artificial intelligence-based tool, artificial neural network (ANN), offers an alternative to RSM-based regression methods. Objectives: This study was conducted to investigate the combined application of experimental design and a neural computing framework for modeling and optimization of a bilayer tablet with biphasic release of cinnarizine (CNZ) and domperidone (DOM), followed by in-silico physiologically-based pharmacokinetic (PBPK) modeling. Methods: The experimental data from the trial formulations supported by the central composite design (CCD) were trained using an artificial neural network. The predicted values of the input variables (HPMC K4M, sodium carbonate, croscarmellose and magnesium stearate) targeting the output responses (% drug release at 1h, 6h, 12h, and friability) were cross-validated using the numerical and graphical optimization technique of CCD. The in-silico PBPK modeling was used to measure relative bioavailability and simulate in-vivo plasma profiles under fasting state through GastroPlus® software. Results: The optimum quantities for developing a bilayer tablet —15% HMPC K4M, 3% sodium carbonate, 2% croscarmellose, and 1% magnesium stearate — were found to be very similar by both the CCD and ANN models, with desirability values close to 1. Moreover, ANOVA revealed no statistically significant difference between the optimized and predicted formulations. GastroPlus® assisted in the relative bioavailability evaluation of the optimized bilayer formulation, with an immediate-release domperidone and an extended-release cinnarizine layer, showing 89% and 81%, respectively. Conclusion: It is concluded that AI-powered modeling, especially the integration of ANN, accelerates innovation, leading to faster and smarter optimization of pharmaceutical formulations.