BLDB : Beta-Lactamase DataBase

We have developed the Beta-Lactamase DataBase (BLDB) in order to overcome the lack of structured and easily accessible information in the field of antibiotic resistance, and more specifically at the level of β-lactamases.

The main feature of this database is the presence in one place of several essential components for a better understanding of the structure-activity relationship of this family of enzymes. The database contains a complete and up-to-date list of all β-lactamases (sequences and three-dimensional structures) described in the literature, as well as a number of mutants and hydrolysis profiles of β-lactam antibiotics. A BLAST-type search functionality is also present. The BLDB is currently used on a daily basis by microbiologists at several antibiotic resistance reference and surveillance centres around the world in their work to identify new variants.

The contents of this database represent a very important element in our work to design effective β-lactamase inhibitors. Currently BLDB contains over 8000 unique enzymes belonging to the 4 classes of β-lactamases, as well as over 1500 crystallographic structures of β-lactamases that are currently available in the Protein Data Bank (PDB). BLDB also contains a limited number of synthetic mutants and hydrolysis profiles. New structures are added semi-automatically every week, after each PDB update. New enzymes are added after each update of the Bacterial Antimicrobial Resistance Reference Gene Database, maintained by NCBI. Constant monitoring of the literature also enables us to identify new β-lactamases and synthetic mutants described in the literature, as well as their hydrolysis profiles.

MDPOW : Calculation of water/solvent partition coefficients with Gromacs

The MDPOW software has been developed in Python, in collaboration with Oliver Beckstein (Arizona State University, USA), and is used to set up and analyse molecular dynamics simulations: an equilibrium simulation (15 ns) followed by 21 simulations of 5 ns each for the perturbation of the free energy. The simulations were carried out with Gromacs in the OPLS-AA force field. The initial version allowed the prediction of hydration free energy (ΔGhyd). We subsequently implemented and validated the prediction of partition coefficients, initially octanol/water (logKOW), then cyclohexane/water (logKCW). We recently improved the reproducibility of the predictions by implementing an NPT protocol instead of NVT.

Although it requires relatively large computational resources, this approach makes it possible to achieve more accurate predictions than the empirical methods currently available.

Conference proceedings :

  • Kenney, I. ; Iorga, B. I. ; Beckstein, O. Towards an automated pipeline for solvation free energy calculations, Annual Meeting of the APS Four Corners Section, Tempe, AZ, United States, Tempe, AZ, United States, 2015.

MOL2FF: Automated parameterisation of ligands in the OPLS-AA force field

Current methods for generating topologies for the OPLS-AA force field all take a similar approach to the AMBER force field, with different partial atomic charges for each molecule that have to be calculated on a case-by-case basis. However, OPLS-AA is much richer than the other force fields (675 types of atoms compared with around 47 for AMBER) and this approach leads to incorrect topologies.

In collaboration with Oliver Beckstein (Arizona State University, USA), we have started to develop a different approach to automated ligand parameterisation, based on chemical function recognition, which enables OPLS-AA atom types to be correctly assigned in all cases. The topologies generated in this way are used to feed the Ligandbook online repository that we recently set up.

The MOL2FF parametrisation software is written in TCL, while the chemical information is stored in SMARTS format and used for substructure searches using the CACTVS software. MOL2FF is currently under development, but an experimental version has already been in use within the team for several years. In particular, it has been used to parameterise the ligands used in the SAMPL3 (2011), SAMPL4 (2013) and SAMPL5 (2015) prediction competitions, as well as the vast majority of compounds present in the Ligandbook online repository. The current version of the software allows the parameterisation of all the chemical functions present in the OPLS-AA force field implemented in Gromacs, as well as some additional heterocycles and chemical functions that have been parameterised in our recent work.

In addition, this method makes it possible to validate the parametrisation in the OPLS-AA force field of compounds for which the values of ΔGhyd and logKSW are known. Good quality OPLS-AA topologies are essential for drug design studies using molecular dynamics. In collaboration with Oliver Beckstein (Arizona State University), we have developed a protocol for testing the quality of OPLS-AA topologies using molecular dynamics (free energy calculations).

Ligandbook: Ligand topologies and force field parameters

The Ligandbook online repository for ligand topologies and force field parameters is a tool that we have developed for sharing information within the modelling community. In this site we have implemented advanced features (search by substructure, automatic processing of deposited structures, links to databases e.g. PDB, PubChem, Common Chemistry, etc.) that will make it much easier to use.

The repository currently contains more than 2,900 structures, the vast majority of ligands from the Protein Data Bank (PDB) and parameterised with our MOL2FF software, and we plan to increase the number of structures by a factor of 10 over the next few years. We hope that Ligandbook will become the reference site in molecular modelling for depositing and downloading ligand topologies and force field parameters, which are essential for the molecular dynamics study of protein-ligand complexes.

Ligandbook is written in PHP in the framework provided by Symfony2 software. The object annotations are stored in a MySQL relational database, while the dropped files are stored directly on disk for better performance. The PHP code is interfaced with TCL scripts run internally by the CACTVS software for the chemoinformatics part. The chemical structures for searches by exact structure or by substructure are drawn using CACTVS Sketcher. Text searches use the Elasticsearch engine. Indexes are created in real time for the vast majority of fields (e.g. IUPAC name, synonyms, chemical formula, molar mass, charge, number of atoms, SMILES, PubChem CID, PDB ligand ID, CAS RN). Indexing these fields enables rapid filtering and searching in Ligandbook.

Articles :

  • Domański, J. ; Beckstein, O. ; Iorga, B. I., Ligandbook – an online repository for small and drug-like molecule force field parameters. Bioinformatics 2017, 33, 1747-1749. [Online version] (Open Access).

Conference proceedings :

  • Domański, J. ; Beckstein, O. ; Iorga, B. I. Ligandbook – an online repository for small and drug-like molecule force field parameters, 244th ACS National Meeting & Exposition Abstracts of Papers, Philadelphia, PA, United States, August 19-23 ; American Chemical Society : Philadelphia, PA, United States, 2012 ; pp COMP-227.