Polymeric drug formulations have significantly improved the safety, efficacy, and clinical impact of many therapies. A persistent challenge for formulation scientists, however, lies in accurately characterizing time-dependent drug release. For decades, researchers have relied on mathematical and physical principles, with a focus on transport phenomena, to interpret release kinetics from various polymeric systems using mechanistic and empirical models. While these models provide a foundational understanding through equations relating diffusion, swelling, and erosion, they often depend on simplifying assumptions and are often limited to a retrospective analysis of in vitro data. Recent advances in artificial intelligence (AI) have since opened the door for a new frontier in modeling strategies. Specifically, machine learning (ML) is being used not only to characterize drug release but also predict it while unveiling key formulation parameters governing unique kinetic profiles. This approach can support faster and more efficient development of polymeric systems. In this review, we explore how traditional drug release models have set the stage for ML in drug delivery research. We discuss important trends across recent ML applications, including data compilation, processing, architecture selection, and performance metrics. This perspective aims to provide scientists with a practical roadmap of ML applications used in formulation development. By integrating these tools with established knowledge, researchers can advance the design and translation of the next generation of polymer-based drug delivery systems. This article is categorized under: Therapeutic Approaches and Drug Discovery > Emerging Technologies.
Woodring et al. (Mon,) studied this question.