Rational development of solubilizing formulations: mechanistic and data-driven approaches to excipient selection

Summary

Background: Pharmaceutical excipients have traditionally been regarded as inactive components of dosage forms. However, in bioenabling formulations, they are essential to achieving sufficient systemic exposure for therapeutic efficacy. Excipients represent a chemically diverse and structurally undefined group. Together with the structural diversity of drugs, this often necessitates trial-and-error approaches during formulation development. For poorly water-soluble drugs (PWSDs), extensive screening campaigns are typically required to identify suitable candidates, consuming substantial time, materials, and resources. A mechanistic understanding of drug-excipient interactions is crucial to overcome barriers to oral drug absorption and to rationalize excipient selection. This is particularly important in preclinical stages, where compound availability and timelines are limited. Computational prediction of excipient selection for lead candidate drugs, together with the design of novel solubilizing excipients, offers the potential to accelerate formulation development while expanding the available toolbox for addressing challenging drug properties

Aim

The aim of this thesis was to advance the understanding and selection of excipients and formulation strategies for PWSDs through mechanistic and data driven approaches. Mechanistically, a focus was placed on the polymeric excipient Soluplus®, used in amorphous solid dispersion (ASD) and preclinical formulations, to further the understanding of its behavior in solution under biorelevant conditions. Its use as a preclinical solubilizer, including potential synergy with ionic surfactants, was further explored to enhance solubilization performance. A second aim was to advance quantitative-structure-property relationships (QSPR) modeling for excipient selection by generating extensive solubility datasets in medium-chain triglyceride (MCT) and aqueous surfactant solutions. These data were used to develop predictive models using machine learning (ML) and to identify structural trends in solubilization performance. By combining mechanistic characterization with data- driven modeling, this work aimed to strengthen the scientific basis for excipient selection and advance a more predictive approach to formulation development.

Methods

In Chapter II, the phase behavior of Soluplus® was investigated under biorelevant conditions using cloud point analysis, multi-angle dynamic light scattering (MADLS), and nuclear magnetic resonance (NMR) spectroscopy. Its effect on drug supersaturation and precipitation was assessed by solvent-shift experiments and dissolution testing of ASDs. Chapter III evaluated the solubilization performance of Soluplus® alone and in combination with ionic surfactants using high-throughput solubility screening across a panel of PWSDs. Additional techniques such as dynamic light scattering (DLS) and viscosity measurements were used to characterize the formulation vehicles. Chapter IV applied ML to develop QSPR models using 182 solubility values in MCT, benchmarking molecular descriptor sets and regression algorithms. Chapter V presented solubility data for 52 drugs in 10 aqueous surfactant systems, comparing solubilization performance through descriptors such as solubility enhancement ratio and molar solubilization capacity.

Results

Soluplus® was characterized under biorelevant conditions for its colloidal properties. The polymer exhibited a miscibility gap dependent on concentration and medium at 37 °C, resulting in coexistence of micelles and polymer-rich coacervate colloids. Drug molecules were shown to associate with the polymer-rich phase, a key consideration when working with Soluplus®-containing media due to observed temperature reversibility. These findings enhance the mechanistic understanding of Soluplus® as a solubilizer or precipitation inhibitor in vitro. Combining Soluplus® with ionic surfactants led to practically relevant solubility increases for most PWSDs investigated, with synergistic effects observed in several cases. Ionic surfactants altered Soluplus®’s aggregation behavior, suggesting a promising strategy for pre-clinical solution formulations when conventional vehicles fail to achieve sufficient drug loading.

Chapter IV has shown how the application of QSPR through ML can aid in rationalizing the selection of formulation approaches by the development of a data driven model that can be applied to quantitatively predict the solubility in MCT. Regularized regression and atom-centered smooth overlap of atomic positions (SOAP) descriptors yielded interpretable and accurate models, revealing how specific structural motifs contribute to solubility. This demonstrated the potential of predictive modeling to guide rational formulation strategy selection. A dataset of 52 drugs across 10 surfactant systems was obtained in Chapter V, which revealed systematic solubilization trends, particularly among structurally related surfactants. Ethoxylated surfactants displayed distinct mechanisms compared to ionic and non-ethoxylated surfactants, indicating that solubilization extends beyond core micellar interactions and depends on specific molecular features of both surfactant and drug.

Conclusion

This thesis integrates mechanistic investigation with data-driven modeling to deepen the understanding of drug–excipient interactions. By elucidating the solution behavior and solubilization mechanisms of Soluplus®, and by applying predictive modeling to excipient selection, it establishes a more systematic foundation for formulation development. The findings demonstrate that combining mechanistic understanding with quantitative prediction enables more rational and efficient design of solubility-enhancing and bioenabling formulations for PWSDs.

Lange, J. J. 2025. Rational development of solubilizing formulations: mechanistic and data-driven approaches to excipient selection. PhD Thesis, University College Cork.