(T1430-10-57) New Developments in the MolMeDB Database of Membrane Interactions of Small Molecules

Purpose: Biological membranes represent a typical barrier for drug transport toward its molecular target. Hence small molecule interactions with biological membranes play a key role in pharmacokinetics and drug delivery. Unfortunately, the data about membrane interactions of small molecules and drugs are in the literature reported in various formats and quality, which effectively hinders the development of this field. For this reason, we developed the MolMeDB database in 2019 [1] (https://molmedb.upol.cz). MolMeDB is a comprehensive and interactive database about interactions of (small) molecules with membranes, including both in-vitro and in-silico permeability data. Expanding the MolMeDB database was to collect a sufficient amount of high-quality data to enable the usage of machine learning techniques that require consistent and plentiful data to enhance their use for drug delivery.

Methods: This process included adding new molecules and their interactions with molecular transporters, biological membranes, and data-validation steps. Data collection into the MolMeDB from the published literature is regularly carried out using an in-house data uploader and a database open source at GitHub (https://github.com/MolMeDB/MolMeDB). Molecules of interest are also calculated in-house by COSMOmic X5 and COSMOperm X8 programs [2]. Transporter data was obtained in collaboration with A. Tuerkova and B. Zdrazil with the usage of their semi-automatic workflow in KNIME [3]. This workflow was based on their previously published workflow [4]. The updated version was able to fetch data about interactions of small molecules with transporters (based on their UniProt IDs) from four databases - IUPHAR Guide to IMMUNOPHARMACOLOGY [5], Metrabase [6], PubChem [7] and ChEMBL [8]. We were mainly focused on ABC and SLC transporters, sodium, potassium and calcium channels and ATPases. Data quality was improved by providing interoperability with other biological databases (Pubchem, ChEMBL, PDB [9]). These databases are used for data validation, mainly for consistent molecular descriptor checks. Each molecule is assigned a unique identificator registered at https://registry.identifiers.org/registry/molmedb.  All membranes, methods and transporters are hierarchically divided by their corresponding properties; nevertheless, dividing molecules is more complicated. MMPA molecule fragmentation system was implemented to provide better insight into molecular properties [10], which are commonly determined by the presence of functional groups in molecule structure.

Results: MolMeDB currently holds data from 1471 publications and in-house COSMOperm calculations, including approximately 950,000 interactions of 450,000 molecules. A few orders of magnitude enlarged from our original publication, which held only 3,600 interactions of 1,200 molecules in 2019. We currently cover 52 methods and 40 membranes, including computational, experimental and simulation on generic or specific membranes. (Graphics 1)  MolMeDB now also holds over 25,800 transporter interactions to 208 targets with the majority of SLC and ABC transporter families and calcium or sodium voltage-gated ion transporters (Graphics 2).  These data allow comparing the membrane behavior for similar molecules, as seen in an example of estrone and estriol (Graphics 3). While both molecules reside in the membrane at similar depths, estrone is more lipophilic and more permeable since it has a smaller barrier for membrane crossing. This type of data allows for the study of functional groups' influence on the permeability and partitioning of small molecules.

Conclusion: MolMeDB data are freely accessible using web UI at the web address https://molmedb.upol.cz/ as CSV downloads of individual parts of the database or via implemented REST API with RDF support. MolMeDB also allows performing comparative analyzes of drug interactions that can be used to guide applications, e.g. for the liposomal formulation of drugs (see our companion poster on In silico tuning of API for liposomal drug delivery).

References: [1] J. Juračka et al., “MolMeDB: Molecules on Membranes Database,” Database, vol. 2019, 2019.
[2] J. A. H. Schwöbel et al., “COSMO perm: Mechanistic Prediction of Passive Membrane Permeability for Neutral Compounds and Ions and Its pH Dependence,” J. Phys. Chem. B, vol. 124, no. 16, pp. 3343–3354, 2020.
[3] M. R. Berthold et al., “KNIME - the Konstanz information miner,” ACM SIGKDD Explor. Newsl., vol. 11, no. 1, pp. 26–31, 2009.
[4] A. Tuerkova et al., “A ligand-based computational drug repurposing pipeline using KNIME and Programmatic Data Access: case studies for rare diseases and COVID-19,” J. Cheminform., vol. 12, no. 1, p. 71, 2020.
[5] J. F. Armstrong et al., “The IUPHAR/BPS Guide to PHARMACOLOGY in 2020: extending immunopharmacology content and introducing the IUPHAR/MMV Guide to MALARIA PHARMACOLOGY,” Nucleic Acids Res., 2019.
[6] L. Mak et al., “Metrabase: A cheminformatics and bioinformatics database for small molecule transporter data analysis and (Q)SAR modeling,” J. Cheminform., vol. 7, no. 1, 2015.
[7] S. Kim et al., “PubChem 2019 update: improved access to chemical data,” Nucleic Acids Res., vol. 47, no. D1, pp. D1102–D1109, 2019.
[8] D. Mendez et al., “ChEMBL: towards direct deposition of bioassay data,” Nucleic Acids Res., vol. 47, no. D1, pp. D930–D940, 2019.
[9] S. K. Burley et al., “Protein Data Bank: the single global archive for 3D macromolecular structure data,” Nucleic Acids Res., vol. 47, no. D1, pp. D520–D528, 2019.
[10] J. Hussain and C. Rea, “Computationally Efficient Algorithm to Identify Matched Molecular Pairs (MMPs) in Large Data Sets,” J. Chem. Inf. Model., vol. 50, no. 3, pp. 339–348, 2010.

Acknowledgments: Authors acknowledge support by Palacky University Olomouc (project IGA_PrF_2022_019 and subproject DSGC-2021-0060 within OP VVV CZ.02.2.69/0.0/0.0/19_073/0016713) and ELIXIR-CZ infrastructure (project LM2018131).

Graphic 3 - Difference in membrane interactions of estrone and estriol as shown by the MolMeDB database. While both molecules reside in the membrane at similar depths, estrone is more lipophilic and more permeable since it has a smaller barrier for membrane crossing.



Graphic 2 - Transporters and ion channel targets in MolMeDB.



Graphic 1 - Content of MolMeDB database in July 2022 divided by methods (left image) and membranes (right image) sorted by the method type and membrane localisation.