Alžběta Türková

Title of the Doctoral Thesis: Data-driven Molecular Modeling Studies with a Special Focus on Hepatocellular Uptake Transporters

Publishing year: 2021

Tags: Organic Anion Transporting Polypeptides / Solute Carrier Transporters / Major Facilitator Superfamily / Organic Cation Transporter 1 / Selectivity Profiling / Molecular Docking / Data Science / Data Integration / Normal Mode Analysis / Structure-based modeling


Uptake transporters of the solute carrier (SLC) superfamily expressed in hepatocytes help maintain cellular homeostasis by regulating the transport of both endogenous substrates and xenobiotic compounds. An impaired function of these transporters might lead to clinically relevant drug-drug interactions with a potential development of drug-induced liver injury. In this thesis, we followed a holistic in silico approach, connecting structure-based modeling with data-science (cheminformatics) approaches to gain an in-depth understanding of transporter-ligand interactions. The biological focus of the thesis lied on hepatic organic anion transporting polypeptides belonging to the SLCO family - OATP1B1, OATP1B3, and OATP2B1. In addition, we studied structural determinants of organic cation transporter 1 (OCT1) as another representative of the pharmaceutically relevant hepatic transporters. The research done in the framework of this thesis is structured into six individual studies (Study 1-6). In Study 1, integrative data mining of OATP bioactivities from public databases was performed. Our intention was to analyze the substrate and inhibitor data with respect to data coverage, distribution of bioactivities, and to uncover enriched chemical substructures with pronounced selectivity profiles. Further, binary classification models were developed to identify important molecular features which were used to study commonalities and differences across the three OATPs. In Study 2, R-group decomposition of a congeneric series of analogs derived from 13-epiestrones was done to study the effect of different substituents on OATP2B1 inhibition. Presence of halogenated substituents at the R-2 position was identified as an important molecular determinant of OATP2B1 inhibition. In Study 3, we focused more on the methodological aspects of the thesis and developed an automated modeling pipeline for performing another emerging cheminformatic technique - ligand-based drug repurposing. The usefulness of the workflow was demonstrated for two case studies (GLUT1-deficiency syndrome and COVID-19). In Study 4 we used structure-based modeling to shed light on differences in uptake of clinical substrates between human and mouse hepatic OCT1. Computational modeling attributed the differences between human and mouse OCT1 to hydrophobic packing interactions between TMH1 and TMH2. Study 5 involved signature dynamics of major facilitator superfamily proteins by normal mode analysis, generation of structural models for OATP1B1, OATP1B3, and OATP2B1 by ensemble docking, and the elucidation of binding mode hypotheses for compound derivatives possessing a steroidal scaffold. Differences in binding of steroids to the three transporters were attributed to different electrostatics and shape complementarity. In addition, several non-conserved residues in the N-terminal region provided structural insights into selectivity switches across the three transporters. Study 6 presents novel OATP inhibitors identified upon a combination of different computational approaches (structure-based virtual screening, conformal prediction, proteochemometric and deep learning models, respectively), which were subsequently validated by a transporter inhibition assay. By investigating binding modes of newly identified inhibitors we showed that the differences in the inner cavity across the three transporters were affected by different localization of aromatic residues. In a biological context, the presented thesis ultimately contributed to the elucidation of molecular determinants of hepatic uptake transporters with a special focus on hepatic OATP-ligand interactions and selectivity. Last but not least, leveraging open data using (semi-)automated workflows was found to be a useful approach to increase the confidence of structure-based modeling approaches applied herein.