Designing Optimum Drug Delivery Systems Using Machine Learning Approaches: a Prototype Study of Niosomes

The project introduces a computer‑based framework that uses machine learning to improve niosomal drug delivery systems. First, we collected data from past studies through a systematic literature search and built a dataset of many niosome formulations. The dataset includes eleven input variables that describe the physicochemical properties of the drugs and the formulation ingredients. These inputs are linked to two key quality attributes: particle size and the percentage of drug that is entrapped. We created an artificial neural network (ANN) that uses a hyperbolic tangent sigmoid activation function and Levenberg–Marquardt back‑propagation to capture the complex, non‑linear relationships between inputs and outputs. A sensitivity analysis was added to pinpoint the most

influential formulation factors, which turned out to be the drug‑to‑lipid ratio and the cholesterol‑to‑surfactant ratio. To test how well the global neural network predicts real results, we prepared experimental batches of Donepezil hydrochloride‑loaded niosomes using a factorial design. We then compared the ANN’s predictions with those from the traditional local response surface methodology (RSM). The comparison shows that the neural network provides more accurate and more generalizable predictions for nanocarrier formulation design. Overall, this work provides a clear guide for using machine‑learning tools in pharmaceutical product development.

Message: Hi FE, I need help with “Designing Optimum Drug Delivery Systems Using Machine Learning Approaches: a Prototype Study of Niosomes” in “Biotechnology & Biomedical Engineering”