A predictive analytics approach for rational design of nanomedicines with programmable drug release

A critical property of antimicrobial-delivering nanoparticles is their ability to control the release kinetics of their payloads. The release kinetics from these nanomedicine formulations is challenging to model from first principles due to interactions between polymer, drug, and nanoparticle properties and a correspondingly large parameter space of properties.¹ Common data mining approaches either insufficiently capture nonlinear behavior or obscure interpretation of the structure of the model and dataspace. A multilinear modeling approach is applied to nanoparticle release data to capture nonlinear behavior and reveal design relationships between formulations. Methods: Release kinetics were obtained from polyanhydride nanoparticles encapsulating four antibiotic payloads to build a dataset of 71 formulations with varying copolymer chemistry, encapsulated drug type, and drug loading. Principal component analysis (PCA) and variable importance projection (VIP) were used to identify key polymer, drug, and nanoparticle descriptors that correlated with release kinetics and encapsulation efficiency. A graph theory analysis² was used to visualize nanoformulation performance in multilinear space and allowed modeling of release behavior.