Correlating in vitro release performance and formulation characteristics of oral sustained release solid dosage forms (hydrophilic matrices)

The majority of medicines available are taken orally. Even though the manufacture and development of oral formulations of small molecules is arguably the most established branch in formulation science, it still lacks in efficiency and predictability. The formulation of new drugs requires experimental iteration and testing. While Design of Experiment (DoE) approaches have accelerated these processes, there remains a gap in the ability to predict drug release from hydrophilic matrices as a function of the matrix and drug properties.
Machine learning (ML) approaches have come to the fore in pharmaceutical sciences because of their potential to accelerate drug discovery and formulation development by establishing quantitative structure-property relationships (QSPR) and by providing models to predict which formulation properties are likely to lead to specific drug release characteristics. This has recently been applied to predict drug release from sustained release injectable polymeric particles. [1] For oral formulations with hydrophilic matrices, previously reported ML approaches are limited to 1-2 specific drugs and 1-2 different matrices, formulated using different conditions. [2] There is a need for predictive models that connect the properties of various APIs, formulated in different matrices with their drug release profiles to accelerate the development of new formulations for new drug candidates.
In this project, we propose to develop a machine learning (ML) approach to develop a predictive model that links the physicochemical properties of APIs and those of hydrophilic formulation matrices with the drug release profile from these formulations. We will curate and categorize literature data for small molecules formulated in hydrophilic matrices for oral delivery to develop a database from existing drug release data for hydrophilic matrices and apply ML tools to this database to identify general governing principles and derive a predictive model for drug release from hydrophilic matrices.