Lipoxygenases (LOXs) are enzymes that catalyze formation of the corresponding hydroperoxides from polyunsaturated fatty acids (e. g. linoleic or arachidonic). LOXs are widely expressed in immune, epithelial, and tumor cells, and activation of these enzymes induces numerous structural and metabolic changes. Abnormal LOX activity contributes to a number of pathophysiological conditions including inflammation, skin disorders and tumor genesis.

The 1 388 - compound Enamine lipoxygenase library was designed using two different approaches:

• Docking-based in silico screening, which identified compounds derived from novel scaffolds;
• Similarity search using 2D linear fingerprints and 3D pharmacophore features of the reported lipoxygenase inhibitors, which gave molecules with new side chains.

Library design

Molecular docking

Docking models were built from the PDB reported protein structures 3O8Y, 4NRE and 3D3L. Prior to docking, all structures were optimized and reconstructed to correct gaps and missing side chains (Figure 1). Water molecules coordinated to Fe³⁺ ion were restrained, and their ionization state was assigned. Each docking model contained electrostatic descriptors of the binding site and mandatory positional constraints.

The resulting models were used to screen the MedChem refined subset of ~1.0M molecules of Enamine’s entire stock collection (over 4.6M compounds). Hits were defined as possessing an obligatory alignment for the positional constraints – metal coordination (red) and hydrophobic center (orange) (Figure 2). The docking results were evaluated through comparison of the docking scores and visual inspection of binding poses (presence of H-bonds, degree of exposure to water), which gave 950 compounds (Figure 3).

Similarity search

• Search of ChEMBL database against the lipoxygenase targets gave a reference set of 322 active compounds for three lipoxygenase types (A5, A12, A15). These compounds were further filtered using potency values cut-off (IC₅₀ ≤ 1μM) and MedChem toxicophore filters.
• 2D similarity search with a 0.85 threshold was done using linear fingerprints and Tanimoto algorithm.
• Nearly 500 compounds were selected from Enamine’s screening collection and added to the LOX library.

Examples of the molecules in the library

Z57167646

CHEMBL381149
IC₅₀= 920 nM, 15-hLOX

CHEMBL1571456
IC₅₀= 120 nM, 15-hLOX

Z1213743775

Z45929261

CHEMBL518103
IC₅₀= 40 nM, 5-AhLOX

CHEMBL3604141
IC₅₀= 92 nM, 15-AhLOX

Z2001236270

The kynurenine pathway is a metabolic pathway leading to the production of nicotinamide adenine dinucleotide (NAD+) from the degradation of tryptophan. Disruption in the pathway is associated with wide range of diseases and disorders including infectious diseases (e.g. HIV), neurological disorders (Alzheimer’s disease (AD), Huntington’s disease (HD) and ALS), affective disorders (schizophrenia, depression and anxiety), autoimmune diseases, peripheral conditions and malignancy. The aim of our work was to conduct searches for new potential active compounds for the kynurenine pathway, which, in turn, could be used as convenient and quality starting points for early drug development.

Library design

The search of potential actives was performed using all available protein and ligand structural data for the following targets: indoleamine dioxygenase (IDO), tryptophan dioxygenase (TDO), 3-hydroxyanthranilic acid dioxygenase (3-HAO), kynurenine aminotransferases (KATs), kynurenine 3-monooxygenase (KMO).

Fig. 1. The kynurenine pathway of tryptophan degradation in mammals.

Screening Models preparation and validation

All available 3D structures of the targets were retrieved from PDB. Alternate superposition and comparison of the protein structures showed high sequence identity in every separate protein target > 90 % with RMSD in a range 0.4 - 0.71 Ų. Considering all representative structures only one has been selected for model preparation taking into account ligand binding parameters (protein- native ligand) and geometric parameters of the binding sites.

Fig. 2.Binding site of IDO1 (2D0T (top), 4PK5 (bottom) and example two main subpockets (in blue and yellow) used for molecular docking).

Molecular docking & Pharmacophore search

Basing on compounds with known activity and “protein–native ligand” complexes, ligand-based (IDO: 3, TDO: 4, KATs: 3, KMO: 2) and structure-based (IDO: 3, TDO: 6, 3-HAO: 2, KATs: 3, KMO: 4) pharmacophore models were build and validated. Drug-like Enamine database of over 1M compounds were then screened.

Fig. 3. Results of molecular docking: superposition of native (grey) and hit ligands (yellow) with representation of structure-based pharmacophore features.

Examples of the molecules

Fig. 4. Ligand-based Pharmacophore models used for in silico screening.

Number of compounds in targeted libraries after molecular docking calculation and pharmacophore searches:

Careful analyzes of proteins and their ligand interactions involved in kynurenine pathway were performed. The iterative in silico searches using all available state-of-the-art methodologies were curried-out to yield unique sets of potentially active molecules. Chemotype control was used to enrich the libraries with structural motifs of true positives and new “patent-free” structural cores.

Developed targeted sets are intended for high-probability initial hit discovery in one step for any particular target.

Development of efficient kinase inhibitors has been a long-standing challenge in drug development. The structural and mechanistic characterization of kinase inhibitor provides new strategies to develop specific kinase inhibitors by targeting a binding pocket adjacent or not adjacent to the ATP binding pocket.

Allosteric kinase inhibition is among the most promising and sensitive deactivation mechanism of kinase activity. Four type of possible allosteric kinase pockets were determined by now:

1. Myristoyl pocket;
2. Inhibitor binding mode that occupies part of binding pocket adjacent to the ATP-site;
3. PIF-pocket (regulatory site targeted);
4. Pocket type I^1/2 relatively allosteric regulators of kinases.

All available PDB structures with kinase allosteric inhibitors were analyzed. Most representative structures were used for molecular docking calculation. General algorithm of library design represented on the scheme below:

Examples of the molecules

Fig. 1. Example of molecular docking result into the Myristoyl pocket (4aw1 PDB entry). Native ligand represented in grey sticks, docked ligand – in yellow.

Fig. 2. Example of molecular docking into the ATP-kinase subpocket (3e8n PDB entry). Docking calculations were performed with CoA features.

Fig. 3. Example of molecular docking into the PIF pocket (4aw1 PDB entry).

Fig. 4. Example of molecular docking into the PIF I^1/2 binding site (4aw1). Hit binding mode.

Phenotypic screening has proven its efficacy in drug discovery and become more and more popular approach in search of new actives. Recently researchers have concluded that sharp focused approaches such as target-based screening are useful but may also limit the breadth of new findings.

Development of new phenotypic screening techniques within implementation of specially designed compound libraries makes this approach a powerful tool in repurposing, finding of new mechanism of action, investigation of signaling pathways and discovery of new biological targets. For successful screening campaign special attention must be paid to the source of chemical entities and their annotation. This requires access to richer, diverse and cross-linked biological data associated to chemical compounds. In order to meet all requirements, we designed special compound library intended for phenotypic screens.

The Library has been pre-plated for most convenient access and prompt delivery in various customized formats. Most popular library formats available for immediate supply are summarized in the table below:

Library Design

To create multipurpose phenotypic library we investigated an optimal balance between diversity of biological activities versus structural diversity of small molecules. The library includes 900+ approved drugs and most similar compounds with identified mechanism of action (T>85%, linear fingerprints). In addition, PSL is enriched with 2000+ of annotated potent inhibitors and their biosimilars, covering a broad diversity of biological targets. Compounds from PSL are cell-permeable and possess pharmacology-compliant physicochemical properties.

• Approved drugs and most similar compounds with identified mechanism of action, over 2 000 molecules, that were found from Enamine Collection (T>87%, linear fingerprints).
• Potent inhibitors or highly similar compounds to diversified protein targets, about 5 000 molecules.

Supporting biological data/documentation:

• Provides easy access and analyze information from in-fields.
• Covers different aspects: polypharmacology, number of targets and description.
• Numbers and names of associated diseases are indicated for majority of compounds.

Figure below demonstrates chemical space distribution using 2D prop parameters (ClogP, fraction of Rotatable bonds, donor/acceptor count, etc.), normalized PMI vectors and distribution between classes of the most potent targets, total number of targets per compound and first level of Anatomical Therapeutic Chemical (ATC) Classification.

Our Phenotypic Library can be applied for screening against different protein classes and diseases, affecting adjacent tissues or individual body systems.