Featured Paper: Predicting the Tabletability of Binary Mixtures from Individual Powder Compaction Behavior (International Journal of Pharmaceutics, 2026. DOI:10.1016/j.ijpharm.2026.126690)
Background
Developing a direct compression tablet often requires extensive trial-and-error to identify formulations with adequate mechanical strength. In this study, Ghijs and colleagues developed a machine learning framework that predicts the tabletability of binary powder mixtures using only the properties of the individual components.
Key Findings
Rather than relying solely on empirical mixing rules, the model combines particle properties with fundamental compaction descriptors, including work of compaction, elastic recovery, and in-die porosity. Using a dataset of over 200 formulations containing 33 pharmaceutical materials, the neural network successfully predicted tablet tensile strength across the entire compaction pressure range and consistently outperformed traditional mixing-rule approaches, particularly for poorly compactable APIs.
One notable advantage is that the model requires only 3 to 5 grams of a new material to characterize its compaction behavior, making it attractive for early-stage formulation development when API availability is limited.
Why it matters
This work highlights the growing role of hybrid modeling, where mechanistic pharmaceutical knowledge is combined with machine learning to improve predictive capability. As digital formulation design and AI-assisted pharmaceutical development continue to evolve, approaches like this have the potential to reduce experimental workload, conserve valuable API, and accelerate formulation development.
------------------------------
Tianyi Xiang
PhD Candidate
University of Minnesota
MINNEAPOLIS MN
[email protected]Disclaimer: Opinions expressed are solely my own and do not express the views or opinions of my employer.
------------------------------