Applying Computational Predictions of Biorelevant Solubility Ratio Upon Self-Emulsifying Lipid-Based Formulations...

Computational approaches are increasingly utilised in development of bio-enabling formulations, including self-emulsifying drug delivery systems (SEDDS), facilitating early indicators of success. This study investigated if in silico predictions of drug solubility gain i.e. solubility ratios (SR), after dispersion of a SEDDS in biorelevant media could be predicted from drug properties. Apparent solubility upon dispersion of two SEDDS in FaSSIF was measured for 30 structurally diverse poorly water soluble drugs. Increased drug solubility upon SEDDS dispersion was observed in all cases, with higher SRs observed for cationic and neutral versus anionic drugs at pH 6.5. Molecular descriptors and solid-state properties were used as inputs during partial least squares (PLS) modelling resulting in predictive models for SR_MC (r² = 0.81) and SR_LC (r² = 0.77). Multiple linear regression (MLR) facilitated generation of simplified SR equations with high predictivity (SR_MC r² = 0.74; SR_LC r² = 0.69), requiring only three drug properties; partition coefficient at pH 6.5 (logD_6.5), melting point (Tm) and aromatic bonds as fraction of total bonds (F-AromB). Through using the equations to inform developability classification system (DCS) classes for drugs that have already been licensed as lipid-based formulations, merits for development with SEDDS was predicted for 2/3 drugs.