MultiBacMam Bimolecular Fluorescence Complementation (BiFC) tool-kit identifies new small-molecule inhibitors of the CDK5-p25 protein-protein interaction (PPI).

Protein-protein interactions (PPIs) are at the core of virtually all biological processes in cells. Consequently, targeting PPIs is emerging at the forefront of drug discovery. Cellular assays which closely recapitulate native conditions in vivo are instrumental to understand how small molecule drugs can modulate such interactions. We have integrated MultiBacMam, a baculovirus-based mammalian gene delivery tool we developed, with bimolecular fluorescence complementation (BiFC), giving rise to a highly efficient system for assay development, identification and characterization of PPI modulators. We used our system to analyze compounds impacting on CDK5-p25 PPI, which is implicated in numerous diseases including Alzheimer's. We evaluated our tool-kit with the known inhibitor p5T, and we established a mini-screen to identify compounds that modulate this PPI in dose-response experiments. Finally, we discovered several compounds disrupting CDK5-p25 PPI, which had not been identified by other screening or structure-based methods before.

Functionalized Proline-Rich Peptides Bind the Bacterial Second Messenger c-di-GMP.

c-di-GMP is an attractive target in the fight against bacterial infections since it is a near ubiquitous second messenger that regulates important cellular processes of pathogens, including biofilm formation and virulence. Screening of a combinatorial peptide library enabled the identification of the proline-rich tetrapeptide Gup-Gup-Nap-Arg, which binds c-di-GMP selectively over other nucleotides in water. Computational and CD spectroscopic studies provided a possible binding mode of the complex and enabled the design of a pentapeptide with even higher binding strength towards c-di-GMP. Biological studies showed that the tetrapeptide inhibits biofilm growth by the opportunistic pathogen P. aeruginosa.

Identification of potent and selective cathepsin S inhibitors containing different central cyclic scaffolds.

Starting from the weakly active dual CatS/K inhibitor 5, structure-based design supported by X-ray analysis led to the discovery of the potent and selective (>50,000-fold vs CatK) cyclopentane derivative 22 by exploiting specific ligand-receptor interactions in the S2 pocket of CatS. Changing the central cyclopentane scaffold to the analogous pyrrolidine derivative 57 decreased the enzyme as well as the cell-based activity significantly by 24- and 69-fold, respectively. The most promising scaffold identified was the readily accessible proline derivative (e.g., 79). This compound, with an appealing ligand efficiency (LE) of 0.47, included additional structural modifications binding in the S1 and S3 pockets of CatS, leading to favorable in vitro and in vivo properties. Compound 79 reduced IL-2 production in a transgenic DO10.11 mouse model of antigen presentation in a dose-dependent manner with an ED50 of 5 mg/kg.

BACE1 inhibitors: a head group scan on a series of amides.

A series of amides bearing a variety of amidine head groups was investigated as BACE1 inhibitors with respect to inhibitory activity in a BACE1 enzyme as well as a cell-based assay. Determination of their basicity as well as their properties as substrates of P-glycoprotein revealed that a 2-amino-1,3-oxazine head group would be a suitable starting point for further development of brain penetrating compounds for potential Alzheimer's disease treatment.

ß-Secretase (BACE1) inhibitors with high in vivo efficacy suitable for clinical evaluation in Alzheimer's disease.

An extensive fluorine scan of 1,3-oxazines revealed the power of fluorine(s) to lower the pKa and thereby dramatically change the pharmacological profile of this class of BACE1 inhibitors. The CF3 substituted oxazine 89, a potent and highly brain penetrant BACE1 inhibitor, was able to reduce significantly CSF Aß40 and 42 in rats at oral doses as low as 1 mg/kg. The effect was long lasting, showing a significant reduction of Aß40 and 42 even after 24 h. In contrast to 89, compound 1b lacking the CF3 group was virtually inactive in vivo.

Design of libraries targeting protein-protein interfaces.

TARGETING PPIS: A novel strategy for designing libraries targeting protein-protein interfaces enabled us to identify diverse chemical entry points to interact with therapeutic targets for which conventional screening libraries delivered no or only few hit structures. The concept was experimentally validated by early hit evaluation in biochemical screens and early ADMET profiling.

ParaFrag--an approach for surface-based similarity comparison of molecular fragments.

A frequent task in computer-aided drug design is to identify novel chemotypes similar in activity but structurally different to a given reference structure. Here we report the development of a novel method for atom-independent similarity comparison of molecular fragments (substructures of drug-like molecules). The fragments are characterized by their local surface properties coded in the form of 3D pharmacophores. As surface properties, we used the electrostatic potential (MEP), the local ionization energy (IE(L)), local electron affinity (EA(L)) and local polarizability (POL) calculated on isodensity surfaces. A molecular fragment can then be represented by a minimal set of extremes for each surface property. We defined a tolerance sphere for each of these extremes, thus allowing us to assess the similarity of fragments in an analogous manner to classical pharmacophore comparison. As a first application of this method we focused on comparing rigid fragments suitable for scaffold hopping. A retrospective analysis of successful scaffold hopping reported for Factor Xa inhibitors [Wood MR et al (2006) J Med Chem 49:1231] showed that our method performs well where atom-based similarity metrics fail.

Recent developments in de novo design and scaffold hopping.

This review covers the developments in the fields of de novo ligand design and scaffold hopping since 2006. De novo ligand design was introduced in 1991 as a purely structure-based method to suggest ligands for synthesis and was later augmented by ligand-based approaches. Both structure-based and ligand-based methods identify pharmacophores, as well as shape constraints, and subsequently match these with complementary features embedded into small-molecule topologies. Recently, significant attention has been paid to de novo ligand design in combination with biophysical fragment screening and X-ray structure elucidation. Scaffold hopping has evolved from a niche application of de novo design into a rapidly expanding suite of different software tools, which are used extensively in the pharmaceutical industry.

Guinea pig chymase is leucine-specific: a novel example of functional plasticity in the chymase/granzyme family of serine peptidases.

To explore guinea pigs as models of chymase biology, we cloned and expressed the guinea pig ortholog of human chymase. In contrast to rats and mice, guinea pigs appear to express just one chymase, which belongs to the alpha clade, like primate chymases and mouse mast cell protease-5. The guinea pig enzyme autolyzes at Leu residues in the loop where human chymase autolyzes at Phe. In addition, guinea pig alpha-chymase selects P1 Leu in a combinatorial peptide library and cleaves Ala-Ala-Pro-Leu-4-nitroanilide but has negligible activity toward substrates with P1 Phe and does not cleave angiotensin I. This contrasts with human chymase, which cleaves after Phe or Tyr, prefers P1 Phe in peptidyl 4-nitroanilides, and avidly hydrolyzes angiotensin I at Phe8 to generate bioactive angiotensin II. The guinea pig enzyme also is inactivated more effectively by alpha1-antichymotrypsin, which features P1 Leu in the reactive loop. Unlike mouse, rat, and hamster alpha-chymases, guinea pig chymase lacks elastase-like preference for P1 Val or Ala. Partially humanized A216G guinea pig chymase acquires human-like P1 Phe- and angiotensin-cleaving capacity. Molecular models suggest that the wild type active site is crowded by the Ala216 side chain, which potentially blocks access by bulky P1 aromatic residues. On the other hand, the guinea pig pocket is deeper than in Val-selective chymases, explaining the preference for the longer aliphatic side chain of Leu. These findings are evidence that chymase-like peptidase specificity is sensitive to small changes in structure and provide the first example of a vertebrate Leu-selective peptidase.

Substituted 2-oxo-azepane derivatives are potent, orally active gamma-secretase inhibitors.

A hydroxamic acid screening hit 1 was elaborated to 5,5-dimethyl-2-oxoazepane derivatives exhibiting low nanomolar inhibition of gamma-secretase, a key proteolytic enzyme involved in Alzheimer's disease. Early ADME data showed a high metabolic clearance for the geminal dimethyl analogs which could be overcome by replacement with the bioisosteric geminal difluoro group. Synthesis and structure-activity relationship are discussed and in vivo active compounds are presented.

From astemizole to a novel hit series of small-molecule somatostatin 5 receptor antagonists via GPCR affinity profiling.

The H1R antagonist astemizole was identified as a somatostatin 5 (SST5) receptor antagonist by a comparative sequence analysis of the consensus drug binding pocket of GPCRs. Subsequently, a similarity analysis of GPCR affinity profiles of astemizole versus a set of in-house GPCR-biased combinatorial libraries revealed new chemical entry points that led to a second lead series with nanomolar binding affinity.

Chemical fragment spaces for de novo design.

Chemical fragment spaces are combinations of molecular fragments and connection rules. They offer the possibility to encode an enormously large number of chemical structures in a very compact format. Fragment spaces are useful both in similarity-based (2D) and structure-based (3D) de novo design applications. We present disconnection and filtering rules leading to several thousand unique, medium size fragments when applied to databases of druglike molecules. We evaluate alternative strategies to select subsets of these fragments, with the aim of maximizing the coverage of known druglike chemical space with a strongly reduced set of fragments. For these evaluations, we use the Ftrees fragment space method. We assess a diversity-oriented selection method based on maximum common substructures and a method biased toward high frequency of occurrence of fragments and find that they are complementary to each other.

Prediction of UV and ESI-MS signal intensities.

All major pharmaceutical companies maintain large collections of compounds that are used either for screening against biological targets or as synthetic precursors. The quality assessment of these compounds is typically done by liquid chromatography combined with mass spectroscopy (LC/MS) and UV purity control. To facilitate the analysis of the analytical data, we have built computational models to predict UV and MS signal intensities under experimental LC/MS conditions. The discriminant partial-least-squares technique was used for classifying compounds into those most likely to yield a MS signal and others where the signal is below the detection limit (94% and 88% correct predictions, respectively). In the case of UV prediction, we compared this statistical linear-regression technique to a knowledge-based approach. A combination of both techniques proved to be the most reliable (96/98% correct predictions of UV-active/ UV-inactive compounds). Both models have been incorporated into the automated compound integrity profiling at F. Hoffmann-La Roche.

Database clustering with a combination of fingerprint and maximum common substructure methods.

We present an efficient method to cluster large chemical databases in a stepwise manner. Databases are first clustered with an extended exclusion sphere algorithm based on Tanimoto coefficients calculated from Daylight fingerprints. Substructures are then extracted from clusters by iterative application of a maximum common substructure algorithm. Clusters with common substructures are merged through a second application of an exclusion sphere algorithm. In a separate step, singletons are compared to cluster substructures and added to a cluster if similarity is sufficiently high. The method identifies tight clusters with conserved substructures and generates singletons only if structures are truly distinct from all other library members. The method has successfully been applied to identify the most frequently occurring scaffolds in databases, for the selection of analogues of screening hits and in the prioritization of chemical libraries offered by commercial vendors.

A robust clustering method for chemical structures.

A clustering method based on finding the largest set of disconnected fragments that two chemical compounds have in common is shown to be able to group structures in a way that is ideally suited to medicinal chemistry programs. We describe how markedly improved results can be obtained by using a similarity metric that accounts not just for the size of the shared fragments but also on their relative arrangement in the two parent compounds. The use of a physiochemical atom typing scheme is also shown to provide significant contributions. Results from calculations using a test set consisting of actives from nine different important biological target proteins demonstrate the strengths of our clustering method and the advantages over other approaches that are widely used throughout the pharmaceutical industry.

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.

A validation study on the practical use of automated de novo design.

The de novo design program Skelgen has been used to design inhibitor structures for four targets of pharmaceutical interest. The designed structures are compared to modeled binding modes of known inhibitors (i) visually and (ii) by means of a novel similarity measure considering the size and spatial proximity of the maximum common substructure of two small molecules. It is shown that the Skelgen algorithm generates representatives of many inhibitor classes within a very short time and that the new similarity measure is useful for comparing and clustering designed structures. The results demonstrate the necessity of properly defining search constraints in practical applications of de novo design.