A Real-World Perspective on Molecular Design.

We present a series of small molecule drug discovery case studies where computational methods were prospectively employed to impact Roche research projects, with the aim of highlighting those methods that provide real added value. Our brief accounts encompass a broad range of methods and techniques applied to a variety of enzymes and receptors. Most of these are based on judicious application of knowledge about molecular conformations and interactions: filling of lipophilic pockets to gain affinity or selectivity, addition of polar substituents, scaffold hopping, transfer of SAR, conformation analysis, and molecular overlays. A case study of sequence-driven focused screening is presented to illustrate how appropriate preprocessing of information enables effective exploitation of prior knowledge. We conclude that qualitative statements enabling chemists to focus on promising regions of chemical space are often more impactful than quantitative prediction.

Novel Triazolopyrimidine-Derived Cannabinoid Receptor 2 Agonists as Potential Treatment for Inflammatory Kidney Diseases.

The cannabinoid receptor 2 (CB2) system is described to modulate various pathological conditions, including inflammation and fibrosis. A series of new heterocyclic small-molecule CB2 receptor agonists were identified from a high-throughput screen. Lead optimization gave access to novel, highly potent, and selective (over CB1) triazolopyrimidine derivatives. A preliminary structure-activity relationship was established, and physicochemical properties in this compound class were significantly improved toward better solubility, lipophilicity, and microsomal stability. An optimized triazolopyrimidine derivative, (3S)-1-[5-tert-butyl-3-[(1-cyclopropyltetrazol-5-yl)methyl]triazolo[4,5-d]pyrimidin-7-yl]pyrrolidin-3-ol (39), was tested in a kidney ischemia-reperfusion model, in which it showed efficacy at a dose of 10 mg kg(-1) (p.o.). A significant depletion of the three measured kidney markers indicated a protective role of CB2 receptor activation toward inflammatory kidney damage. Compound 39 was also protective in a model of renal fibrosis. Oral treatment with 39 at 3 mg kg(-1) per day significantly decreased the amount of fibrosis by ∼ 40% which was induced by unilateral ureter obstruction.

Challenges and Rewards in Medicinal Chemistry Targeting Cardiovascular and Metabolic Diseases.

Medicinal chemistry has been transformed by major technological and conceptual innovations over the last three decades: structural biology and bioinformatics, structure and property based molecular design, the concepts of multidimensional optimization (MDO), in silico and experimental high-throughput molecular property analysis. The novel technologies advanced gradually and in synergy with biology and Roche has been at the forefront. Applications in drug discovery programs towards new medicines in cardiovascular and metabolic diseases are highlighted to show impact and advancement: the early discovery of endothelin antagonists for endothelial dysfunction (Bosentan), 11-beta hydroxysteroid dehydrogenase (11ß-HSD1) inhibitors for dysregulated cellular glucocorticoid tonus (type 2 diabetes and metabolic syndrome) and non-covalent hormone sensitive lipase (HSL) inhibitors to study the scope of direct inhibition of lipolysis in the conceptual frame of lipotoxicity and type 2 diabetes.

Challenges and Rewards in Medicinal Chemistry Targeting Cardiovascular and Metabolic Diseases.

Medicinal chemistry has been transformed by major technological and conceptual innovations over the last three decades: structural biology and bioinformatics, structure and property based molecular design, the concepts of multidimensional optimization (MDO), in silico and experimental high-throughput molecular property analysis. The novel technologies advanced gradually and in synergy with biology and Roche has been at the forefront. Applications in drug discovery programs towards new medicines in cardiovascular and metabolic diseases are highlighted to show impact and advancement: the early discovery of endothelin antagonists for endothelial dysfunction (Bosentan), 11-beta hydroxysteroid dehydrogenase (11ß-HSD1) inhibitors for dysregulated cellular glucocorticoid tonus (type 2 diabetes and metabolic syndrome) and non-covalent hormone sensitive lipase (HSL) inhibitors to study the scope of direct inhibition of lipolysis in the conceptual frame of lipotoxicity and type 2 diabetes.

3-Amido-3-aryl-piperidines: A Novel Class of Potent, Selective, and Orally Active GlyT1 Inhibitors.

3-Amido-3-aryl-piperidines were discovered as a novel structural class of GlyT1 inhibitors. The structure-activity relationship, which was developed, led to the identification of highly potent compounds exhibiting excellent selectivity against the GlyT2 isoform, drug-like properties, and in vivo activity after oral administration.

CONFECT: conformations from an expert collection of torsion patterns.

The generation of sets of low-energy conformations for a given molecule is a central task in drug design. Herein we present a new conformation generator called CONFECT that builds on our previously published library of torsion patterns. Conformations are generated as well as ranked by means of normalized frequency distributions derived from the Cambridge Structural Database (CSD). Following an incremental construction approach, conformations are selected from a systematic enumeration within energetic boundaries. The new tool is benchmarked in several different ways, indicating that it allows the efficient generation of high-quality conformation ensembles. These ensembles are smaller than those produced by state-of-the-art tools, yet they effectively cover conformational space.

Isothermal titration calorimetry with micelles: Thermodynamics of inhibitor binding to carnitine palmitoyltransferase 2 membrane protein.

Carnitine palmitoyl transferase 2 (CPT-2) is a key enzyme in the mitochondrial fatty acid metabolism. The active site is comprised of a Y-shaped tunnel with distinct binding sites for the substrate acylcarnitine and the cofactor CoA. We investigated the thermodynamics of binding of four inhibitors directed against either the CoA or the acylcarnitine binding sites using isothermal titration calorimetry (ITC). CPT-2 is a monotopic membrane protein and was solubilized by ß-octylglucoside (ß-OG) above its critical micellar concentration (CMC) to perform inhibitor titrations in solutions containing detergent micelles. The CMC of ß-OG in the presence of inhibitors was measured with ITC and small variations were observed. The inhibitors bound to rat CPT-2 (rCPT-2) with 1:1 stoichiometry and the dissociation constants were in the range of K D = 2-20 µM. New X-ray structures and docking models of rCPT-2 in complex with inhibitors enable an analysis of the thermodynamic data in the context of the interaction observed for the individual binding sites of the ligands. For all ligands the binding enthalpy was exothermic, and enthalpy as well as entropy contributed to the binding reaction, with the exception of ST1326 for which binding was solely enthalpy-driven. The substrate analog ST1326 binds to the acylcarnitine binding site and a heat capacity change close to zero suggests a balance of electrostatic and hydrophobic interactions. An excellent correlation of the thermodynamic (ITC) and structural (X-ray crystallography, models) data was observed suggesting that ITC measurements provide valuable information for optimizing inhibitor binding in drug discovery.

Torsion angle preferences in druglike chemical space: a comprehensive guide.

Crystal structure databases offer ample opportunities to derive small molecule conformation preferences, but the derived knowledge is not systematically applied in drug discovery research. We address this gap by a comprehensive and extendable expert system enabling quick assessment of the probability of a given conformation to occur. It is based on a hierarchical system of torsion patterns that cover a large part of druglike chemical space. Each torsion pattern has associated frequency histograms generated from CSD and PDB data and, derived from the histograms, traffic-light rules for frequently observed, rare, and highly unlikely torsion ranges. Structures imported into the corresponding software are annotated according to these rules. We present the concept behind the tree of torsion patterns, the design of an intuitive user interface for the management and usage of the torsion library, and we illustrate how the system helps analyze and understand conformation properties of substructures widely used in medicinal chemistry.

Highly potent and selective cannabinoid receptor 2 agonists: initial hit optimization of an adamantyl hit series identified from high-through-put screening.

A series of highly potent & selective adamantane derived CB2 agonists was identified in a high-throughput screen. A SAR was established and physicochemical properties were significantly improved. This was accompanied by potency of the compounds on the Q63R variant and varying ß-arrestin data which will support the insight into their relevance for the in vivo situation.

TFD: Torsion Fingerprints as a new measure to compare small molecule conformations.

Advantages like intuitive interpretation, objectivity, general applicability, and its easy, automated calculation make the rmsd (root-mean-squared deviation) the measure of choice for the investigation of the accuracy of conformational model generators. For comparing conformations of a single molecule this is a clearly superior method. Single molecule analysis is, however, a rare scenario. Typically, conformations are generated for huge corporate or external vendor databases of high diversity which are then further investigated with high-throughput computational methods like docking or pharmacophore searching, in virtual screening campaigns. Representative subsets for accuracy investigations of computational methods need to mimic this diversity. Averaged rmsd values over these data sets are frequently used to assess the accuracy of the methods. There are, however, significant weaknesses in rmsd comparisons for such kind of data sets. The interpretation is for example no longer intuitive because what can be expected in terms of good or bad rmsd values crucially depends on the data set composition like size or number of rotatable bonds of the underlying molecules. Further, rmsd lacks normalization which might result in very high averaged rmsd values for highly flexible molecules and thus might completely skew results. We have developed a novel measure to compare conformations of molecules called Torsion Fingerprint Deviation (TFD). It extracts, weights, and compares Torsion Fingerprints from a query molecule and generated conformations under consideration of acyclic bonds as well as ring systems. TFD is alignment-free and overcomes major limitations of rmsd while retaining its advantages.

NAOMI: on the almost trivial task of reading molecules from different file formats.

In most cheminformatics workflows, chemical information is stored in files which provide the necessary data for subsequent calculations. The correct interpretation of the file formats is an important prerequisite to obtain meaningful results. Consistent reading of molecules from files, however, is not an easy task. Each file format implicitly represents an underlying chemical model, which has to be taken into consideration when the input data is processed. Additionally, many data sources contain invalid molecules. These have to be identified and either corrected or discarded. We present the chemical file format converter NAOMI, which provides efficient procedures for reliable handling of molecules from the common chemical file formats SDF, MOL2, and SMILES. These procedures are based on a consistent chemical model which has been designed for the appropriate representation of molecules relevant in the context of drug discovery. NAOMI's functionality is tested by round robin file IO exercises with public data sets, which we believe should become a standard test for every cheminformatics tool.

De novo design by pharmacophore-based searches in fragment spaces.

De novo ligand design supports the search for novel molecular scaffolds in medicinal chemistry projects. This search can either be based on structural information of the targeted active site (structure-based approach) or on similarity to known binders (ligand-based approach). In the absence of structural information on the target, pharmacophores provide a way to find topologically novel scaffolds. Fragment spaces have proven to be a valuable source for molecular structures in de novo design that are both diverse and synthetically accessible. They also offer a simple way to formulate custom chemical spaces. We have implemented a new method which stochastically constructs new molecules from fragment spaces under consideration of a three dimensional pharmacophore. The program has been tested on several published pharmacophores and is shown to be able to reproduce scaffold hops from the literature, which resulted in new chemical entities.

Structure-assisted discovery of an aminothiazole derivative as a lead molecule for inhibition of bacterial fatty-acid synthesis.

Fatty-acid synthesis in bacteria is of great interest as a target for the discovery of antibacterial compounds. The addition of a new acetyl moiety to the growing fatty-acid chain, an essential step in this process, is catalyzed by beta-ketoacyl-ACP synthase (KAS). It is inhibited by natural antibiotics such as cerulenin and thiolactomycin; however, these lack the requirements for optimal drug development. Structure-based biophysical screening revealed a novel synthetic small molecule, 2-phenylamino-4-methyl-5-acetylthiazole, that binds to Escherichia coli KAS I with a binding constant of 25 microM as determined by fluorescence titration. A 1.35 A crystal structure of its complex with its target reveals noncovalent interactions with the active-site Cys163 and hydrophobic residues of the fatty-acid binding pocket. The active site is accessible through an open conformation of the Phe392 side chain and no conformational changes are induced at the active site upon ligand binding. This represents a novel binding mode that differs from thiolactomycin or cerulenin interaction. The structural information on the protein-ligand interaction offers strategies for further optimization of this low-molecular-weight compound.

Recore: a fast and versatile method for scaffold hopping based on small molecule crystal structure conformations.

Replacing central elements of known active structures is a common procedure to enter new compound classes. Different computational methods have already been developed to help with this task, varying in the description of possible replacements, the query input, and the similarity measure used. In this paper, a novel approach for scaffold replacement and a corresponding software tool, called Recore, is introduced. In contrast to prior methods, our main objective was to combine the following three properties in one tool: to avoid structures with strained conformations, to enable the exploration of large search spaces, and to allow interactive use through short response times. We introduce a new technique employing 3D fragments generated by combinatorial enumeration of cuts. It allows focusing on fragments suitable for scaffold replacement while retaining conformational information of the corresponding crystal structures. Based on this idea, we present an algorithm utilizing a geometric rank searching approach. Given a geometric arrangement of two or three exit vectors and additional pharmacophore features, the algorithm finds fragments fulfilling all these constraints ordered by increasing deviation from the query constraints. For the validation of the approach, three different design scenarios have been used. The results obtained show that our approach is able to propose new valid scaffold topologies.

Enzyme redesign: two mutations cooperate to convert cycloartenol synthase into an accurate lanosterol synthase.

Efforts to modify the catalytic specificity of enzymes consistently show that it is easier to broaden the substrate or product specificity of an accurate enzyme than to restrict the selectivity of one that is promiscuous. Described herein are experiments in which cycloartenol synthase was redesigned to become a highly accurate lanosterol synthase. Several single mutants have been described that modify the catalytic specificity of cycloartenol to form some lanosterol. Modeling studies were undertaken to identify combinations of mutations that cooperate to decrease the formation of products other than lanosterol. A double mutant was constructed and characterized and was shown to cyclize oxidosqualene accurately to lanosterol (99%). This catalytic change entailed both relocating polarity with a His477Asn mutation and modifying steric constraints with an Ile481Val mutation.

Access of the substrate to the active site of yeast oxidosqualene cyclase: an inhibition and site-directed mutagenesis approach.

A structural model of Saccharomyces cerevisiae oxidosqualene cyclase (SceOSC) suggests that some residues of the conserved sequence Pro-Ala-Glu-Val-Phe-Gly (residues 524-529) belong to a channel constriction that gives access to the active-site cavity. Starting from the SceOSC C457D mutant, which lacks the cysteine residue next to the catalytic Asp456 residue Cys457 has been replaced but Asp456 is still there, we prepared two further mutants where the wild-type residues Ala525 and Glu526 were individually replaced by cysteine. These mutants, especially E526C, were very sensitive to the thiol-reacting agent dodecyl-maleimide. Moreover, both the specific activity and the thermal stability of E526C were severely reduced. A similar decrease of the enzyme functionality was obtained by replacing Glu526 with alanine, while substitution with the conservative residues aspartate or glutamine did not alter catalytic activity. Molecular modeling of the yeast wild-type OSC and mutants on the template structure of human OSC confirms that the channel constriction is an important aspect of the protein structure and suggests a critical structural role for Glu526.

Validation and use of the MM-PBSA approach for drug discovery.

The MM-PBSA approach has become a popular method for calculating binding affinities of biomolecular complexes. Published application examples focus on small test sets and few proteins and, hence, are of limited relevance in assessing the general validity of this method. To further characterize MM-PBSA, we report on a more extensive study involving a large number of ligands and eight different proteins. Our results show that applying the MM-PBSA energy function to a single, relaxed complex structure is an adequate and sometimes more accurate approach than the standard free energy averaging over molecular dynamics snapshots. The use of MM-PBSA on a single structure is shown to be valuable (a) as a postdocking filter in further enriching virtual screening results, (b) as a helpful tool to prioritize de novo design solutions, and (c) for distinguishing between good and weak binders (DeltapIC(50) > or = 2-3), but rarely to reproduce smaller free energy differences.

Insight into steroid scaffold formation from the structure of human oxidosqualene cyclase.

In higher organisms the formation of the steroid scaffold is catalysed exclusively by the membrane-bound oxidosqualene cyclase (OSC; lanosterol synthase). In a highly selective cyclization reaction OSC forms lanosterol with seven chiral centres starting from the linear substrate 2,3-oxidosqualene. Valuable data on the mechanism of the complex cyclization cascade have been collected during the past 50 years using suicide inhibitors, mutagenesis studies and homology modelling. Nevertheless it is still not fully understood how the enzyme catalyses the reaction. Because of the decisive role of OSC in cholesterol biosynthesis it represents a target for the discovery of novel anticholesteraemic drugs that could complement the widely used statins. Here we present two crystal structures of the human membrane protein OSC: the target protein with an inhibitor that showed cholesterol lowering in vivo opens the way for the structure-based design of new OSC inhibitors. The complex with the reaction product lanosterol gives a clear picture of the way in which the enzyme achieves product specificity in this highly exothermic cyclization reaction.

Ensemble methods for classification in cheminformatics.

We describe the application of ensemble methods to binary classification problems on two pharmaceutical compound data sets. Several variants of single and ensembles models of k-nearest neighbors classifiers, support vector machines (SVMs), and single ridge regression models are compared. All methods exhibit robust classification even when more features are given than observations. On two data sets dealing with specific properties of drug-like substances (cytochrome P450 inhibition and "Frequent Hitters", i.e., unspecific protein inhibition), we achieve classification rates above 90%. We are able to reduce the cross-validated misclassification rate for the Frequent Hitters problem by a factor of 2 compared to previous results obtained for the same data set with different modeling techniques.