Prediction of bile salt export pump inhibition using biomimetic chromatographic descriptors: Integrating membrane...

Bile salt export pump (BSEP) inhibition is a key mechanistic driver of cholestatic drug-induced liver injury, yet current prediction approaches rely mainly on in vitro potency and calculated physicochemical descriptors that do not adequately reflect local exposure at the canalicular membrane. This study aimed to evaluate whether biomimetic chromatographic descriptors could improve predictions of BSEP inhibition by incorporating mechanistically relevant information on membrane affinity and protein binding.

A dataset of 62 structurally diverse compounds with reported BSEP inhibition data was compiled and modelled using multiple linear regression. Biomimetic descriptors derived from immobilized artificial membrane chromatography and human serum albumin affinity (log k _HSA) were compared against conventional descriptors such as lipophilicity and molecular weight. Model performance was assessed using internal and external validation, graphical diagnostics, Y-randomization, and applicability domain analysis.

Models based on biomimetic descriptors demonstrated superior, more balanced predictive performance than conventional physicochemical models, with greater robustness and external predictivity. The combination of membrane-affinity and protein-binding descriptors provided a more accurate representation of the determinants governing BSEP interactions.

Biomimetic chromatographic descriptors provide an experimentally grounded approximation of membrane-proximal exposure, a key determinant of BSEP interactions that conventional descriptors do not capture. This approach extends beyond traditional quantitative structure-activity relationship by incorporating distribution-relevant properties into predictive modelling and offers a simple, interpretable framework to support early identification of transporter-mediated toxicity risk in drug discovery.