At BI’s oncology research department in Vienna, we have consequently enabled medicinal chemists to autonomously undertake structure-based compound design and lead optimization. The key remit of computational chemistry is hit finding, enablement and deployment of design tools, and specific project-driven assignments.

A web-based compound design platform, referred to as Chladde, has been engineered, and allows medicinal chemists to devise, handle and share new design ideas. MOE is used as the tool for performing all 3D-structural modelling and design work, and is tighly integrated within the Chladde environment. MOE and Chladde are interfaced with a number of global computational chemistry calculation engines that can be directly invoked by users.

The talk will outline the concepts of this design strategy. The focus will be on the provision of pre-processed structural data and computational tools that are readily available from within the MOE/Chladde environment. Pre-procesed structural data are provided as MOEProject files and automatically enriched with additional project-relevant annotation. Simple graphical tools enable to construct and geometrically analyse design scenarios, and annotate them for sharing. Grid-based approaches facilitate the incorporation of complex map-based computational methods into the design process, and include, for example, the visualisation of ligand snugness-of-fit and void analysis, and MD-derived hot spot maps. Direct connections to molecular property and quantum mechanics based calculation services, and their use in compound design, will also be discussed. New approaches are continuously added to the design platform, expanding the computational compound design toolbox for medicinal chemists.

Cancer Research UK Cancer Therapeutics Unit, Division of Cancer Therapeutics, The Institute of Cancer Research, London, SM2 5NG, UK.

Many medicinal chemistry relevant structures and core scaffolds tend towards geometric planarity. However, structural planarity may preclude the optimisation of physicochemical properties desirable in drug-like molecules, such as solubility. Furthermore, as new and challenging drug targets, such as protein-protein interactions, become more prevalent in drug design projects, the inherent potential to exploit three-dimensionality of chemical structures in lead optimisation is increasingly important. To this end, there has been recent interest in designing molecular fragments for fragment screening and subsequent derivatisation that exhibit enhanced three-dimensionality [1]. However, it remains unclear the extent to which core scaffolds require enhanced three-dimensionality in order to yield molecular designs with desired three-dimensionality.

Here, three computational methods are applied to investigate the emergence of three-dimensionality in drug-like molecules, namely: fragmentation analysis using a recently reported fragmentation algorithm, SynDiR [2]; iterative pruning of pendant substituents using the Scaffold Tree fragmentation rules [3,4]; and the virtual enumeration of drug-like molecules from molecular fragments of varying three-dimensionality. Using the recently published three-dimensionality descriptor Plane of Best Fit (PBF), amongst other descriptors, it is possible to assess the potential three-dimensionality of molecular fragments objectively [5]. The combination of these three approaches to investigate the emergence of three-dimensionality in drug-like molecules informs on the stages at which three-dimensionality should be considered in a drug design project. These methods permit a greater understanding of the properties of the derived functional groups and scaffolds from exemplified medicinal chemistry space and their contributions to three- dimensionality. This study has highlighted key learning that is anticipated to enhance medicinal chemistry design in the future.

References

53 Portland Street, Manchester, M1 3LD, UK
eMail: Thorsten.Nowak@C4XDiscovery.com

Drug design is greatly facilitated by the use of small molecule conformational data, which allows a detailed, mechanistic contextualisation of the binding interaction. While 3D-data for the target protein in both free and bound states is often accessible through X-ray crystallography, experimental 3D-data for the ligand in the unbound solution state is often harder to obtain at high resolution. Since the free ligand’s 3D-shape is an integral part of the binding equation, it also provides crucial information for understanding and controlling the interaction. One technique that excels in providing this information is solution Nuclear Magnetic Resonance (NMR).

Solution NMR offers a highly versatile tool to access information on small molecules under physiologically relevant conditions. Free ligand conformational states in particular can be measured and often readily assessed for changes across a series by diagnostic spectral features that can be easily measured. Even small amounts of easy-to-measure ligand data can greatly aid in silico ligand conformational analysis and SAR interpretation. The availability of experimentally validated ligand conformational data greatly ameliorates docking and pharmacophore searching and provides opportunities for improving the accuracy of analysis, prediction and design. The use of quantitative ligand NMR data can be very useful to contextualise protein crystallographic information and the interpretation of molecular recognition events.

The presentation will demonstrate the use of ligand NMR 3D-data as a vital tool for rational drug design, highlighting its synergy with X-ray crystallography and computational modelling. Taking examples from the literature [1] in addition to ones from our in-house programmes, I will illustrate how solution NMR data from both routine and advanced methods [2] have been used to guide docking studies, the description of pharmacophore data and produce better compounds by rational design.

References

Modern, target-focused drug discovery is typically built around biochemical and cell biological assays. Such assays are employed to find and develop potent small molecule modulators of target activity. Often left out of the process is a means to sensitively and accurately characterize the direct interaction of the small molecules with the target. However, careful implementation of biophysical approaches can help to prevent wasted time chasing artifacts and insure optimal physico-chemical properties of compounds derived from High Throughput Screening (HTS). Moreover, biophysics can be used as the primary means to drive drug discovery as in the Fragment Based approach (FBDD), in which ZoBio is specialized.

Given the diversity of targets in current portfolios, it is critical to have an integrated, flexible discovery pipeline. ZoBio has developed and applied an array of technologies including:

• Protein engineering/optimization for biophysics & structural biology including Domain Trapping technology capable of screening millions of protein variations for the best behaved. Fragment library screening using our proprietary TINS NMR screening technology, Biacore (SPR) and/or High Concentration Screening (HCS). A diverse fragment library of roughly 1,700 compounds, in silico selection of target-specific libraries and screening of customer’s libraries. Orthogonal validation of screening hits. Triaging of HTS compounds with our Biacore 4000 instrument. Parallel application of both X-ray crystallography and our “NMR Pyramid” to provide 3D structural information for structure-based drug design (SBDD).

Combined, these approaches create a powerful, robust pipeline that can be applied to challenging targets to go from a gene to cellular proof-of-concept quickly and efficiently.

All alternative to the method of Group Contributions (atom types) in QSPR modeling is presented. A small number of descriptors derived from a modified Hückel Theory (2D) calculation are used in place of tens or hundreds of atom type descriptors for creating QSPR models of Molar Refractivity, logP, logS, Boiling Point and Free Energy of Hydration. The method is extended to the calculation of pKa and logD. The results of computational experiments demonstrate the validity of the approach.

Literature analyses have come to varying conclusions regarding the physical properties of antibacterial compounds. Our recent analysis of data from Chembl has shown no significant difference in properties such as molecular weight and logP between compounds with and without antibacterial activity.

We have extended this analysis to examine whether the space spanned by InhibOx's ElectroShape descriptors captures any aspects of antibacterial activity. These have previously been successful in the rapid virtual screening of large compound libraries. Their extension to the analysis of antibacterial compounds indicates that some regions of ElectroShape space are enriched in compounds that penetrate bacteria. Tools such as clustering and Self-Organising Maps have been used to identify these regions of space with the aim of guiding the optimisation of lead molecules and improving their antibacterial-likeness.

The identification and design of selective compounds is important for the reduction of unwanted side effects as well as for the development of tool compounds for target validation studies. This is in particular true for therapeutically important protein families that possess conserved folds and have numerous members such as kinases. Due to the increased coverage of profiling data as well as crystal structures, the rational design of selective compounds across the kinome comes into reach. The presentation will cover our efforts to predict off-targets via a machine learning pipeline as well as a novel approach that allows identification of specificity determining subpockets between closely related kinases taking a short list of off-targets as well as a large number of conformations into account. The latter provides an intuitive visualization of kinase specific subpockets and, thus, guidelines for modifying lead compounds.

Evotec (UK) Ltd., 114 Milton Park, Abingdon, Oxfordshire OX14 4SA, United Kingdom
alexander.heifetz@evotec.com

G-Protein Coupled Receptors (GPCRs) have enormous physiological and biomedical importance, being the primary target of a large number of modern drugs. The availability of structural information on the binding site of the targeted GPCR plays a key role in rationalization, efficiency and cost-effectiveness of the drug discovery process. X-ray crystallography, a traditional source of structural information, is not currently feasible for every GPCR or GPCR-ligand complex. This situation significantly limits the ability of crystallography to impact the drug discovery process for GPCR targets in “real-time” and hence there is an urgent need for other practical alternatives.

We uses our hierarchical GPCR modeling protocol (HGMP) to generate a 3D model of GPCR structures and its complexes with small molecules by applying a set of computational methods.² These computational methods includes a large set of unique plugins to refine the GPCR models and exclusive scoring functions like the GPCR-likeness assessment score (GLAS) to evaluate model quality. HGMP is also “armed” with a pairwise protein comparison method (ProS) used to cluster the structural data generated by the HGMP and to distinguish between different functional sub-states. Recently the capabilities of HGMP have been extended by Fragment Molecular Orbital (FMO), quantum mechanical method, to comprehensive exploration of the receptor-ligand interactions. ¹

The models and modeling insights produced by HGMP are then used in SBDD. In our presentation we will demonstrate how HGMP have been integrated with experimental methods and has been successfully applied in drug discovery projects. ³

References

Actin-related protein 2/3 (Arp2/3) complex is a seven subunit ATP-ase that is a key actin cytoskeleton regulator with roles in bacterial pathogenesis and motility of cancer cells. Arp2/3 binds to existing actin filaments and nucleates new filament growth through activation of the complex that proceeds via a large conformational change, which, however, remains largely underexplored. The biological role of the Arp2/3 complex late in development or in adult stages is undiscovered owing to the lack of potent and reversible Arp2/3 inhibitors. Potent, reversible small molecule inhibitors have the potential to become powerful tools to study the Arp2/3 complex in vivo. A small molecule inhibitor (1) of the Arp2/3 complex has been recently discovered and crystallized in complex with Arp2/3. However, the relatively low potency of 1 increases its potential for off target effects in vivo, complicating interpretation of its influence in cell biological studies. Thus, based on the crystal structure, we used molecular docking and free energy perturbation (FEP) calculations of 1 to guide optimization efforts. Binding free energies determined by FEP were found to be in very good agreement with experimental results. This methodology was thus further utilized to guide lead optimization through iterative rounds of calculations, synthesis, and assaying, leading to the discovery of nanomolar inhibitors. In an effort to discover novel scaffolds of Arp2/3 inhibitors, virtual screening was also employed by docking the Maybridge and ZINC databases into the Arp2/3 binding site of 1. Top-scored ligands were post-processed, the final set of compounds was clustered, and 28 exemplars were selected and assayed. Several novel inhibitors of the Arp2/3 complex were identified. Finally, we performed extensive Molecular Dynamics simulations of the inactive and active states of Arp2/3. Our simulations reveal the motions of the different protein subunits during the conformational change of Arp2/3 providing important insights into the mechanism of the activation of the complex. VI-SEEM resources have been used for this project.

Approaches to perform 3D QSAR analysis, such as the well-known Comparative Molecular Field Analysis (CoMFA)method, have been long established in ligand based drug design efforts. Most 3D QSAR methods are based on computing potential energies (or other quantities) on a grid which surrounds a 3D alignment of small molecules with diverse activities against a given biological structure. The grid-point potentials are then correlated with the activitiesusing statistical methods such as partial least squares to produce a QSAR model which can predict the activity of an aligned 3D query molecule. The contributions of each grid point to the model can be visualized by plotting 3D iso-contours of model coefficients on the grid. These 3D ‘fields’ can be used to understand ligand structural components essential for activity and to suggest where ligand modifications can be made. The literature contains a number of 3D QSAR approaches which use different flavours of grid potentials and different methods of statistical fitting to establish correlations between grid-potentials and activity.

Here we present QuaSAR3D, a highly-customizable, module-based 3D QSAR tool for MOE. Based on a given alignment of small molecules, QuaSAR3D offers streamlined modules for grid computation, model building, field visualization and model evaluation. Starting with classic CoMFA, which utilises a charged carbon atom for grid analysis and partial least squares regression with a leave-one out cross validation for model building, the plugin structure of the application allows optional modification and enhancement to include alternative grid types such as Poisson-Boltzmann electrostatics or DFT properties, various model building algorithms such as Comparative Molecular Similarity Index Analysis (CoMSIA) or Field Extrema analysis, as well as different cross validation techniques. This approach offers a straightforward way to explore the large variety of 3D QSAR approaches and a platform to test and develop new ones.

We have created a new paradigm to explore possible solutions to two of the most relevant uncertainties in drug discovery: the structure-energy and the in vitro-in vivo relationships. Our novel approaches rely on first principles and imply the need to re-think and re-learn structure-based drug design and its application to drug discovery. The talk will exemplify several aspects of this new approach and its application to real-world examples covering the use of kinetics, dynamic modeling, solvation and flexibility.