Tuberculosis (TB) is among the most threatening diseases that can cause the next pandemic across the world. It is one of the most rampant infectious diseases owing largely to the methods of transmission and high death ratio. The absence of efficient treatment and the rapid appearance of drug resistance make TB one of the most emerging tasks to resolve in public health. Continuously mutated, Mtb remains one of the most dangerous and difficult-to-treat bacterial diseases. Lack of early diagnosis and the prevalence of resistant strains make TB extremely difficult for timely detection and treatment. Antituberculosis drug development is a high priority to prevent future pandemics. It needs further investment in scientific research and rational drug design to combat the high death rate caused by this disease.

The library is available in pre-plated formats for quick access and is supported with multiple benefits on hit follow-up studies:

• Analogs and hit samples resupply from dry stock of over 4.4 M compounds with immediate QC check.
• MedChem support in hit follow-up and optimization to lead series.
• Same-day ADME tests for design and synthesis prioritization.

Library Design

To design the library and select the most promising molecules, we pay attention to the complexity of Mtb cell-wall and the difficulties of membrane penetration. Despite the absence of certain ruleless and defined filters, we tried to remove all compounds that could have issues with penetration based on the reported data and structure analysis of cell active vs inactive compounds.

Having studied some special features related to the biochemical pathways and morphological properties inherent to Mtb, a group of protein targets has been selected and prioritized to perform in silico screening (vHTS) and select compounds with the alleged activity. Generally, the protein targets in this library can be divided into two subsets: Cell Wall Synthesis-associated proteins and Mtb-specific Essential protein targets distinct from those present in eukaryotic organisms.

Alanine racemase (Alr), which catalyzes the interconversion of L-alanine and D-alanine. Which in turn plays a key role in peptidoglycan cross-linking.

dTDP-4-dehydrorhamnose 3,5-epimerase (RmlC) is involved in the biosynthesis of the dTDP-L-rhamnose which is required for connection of the galactan region of arabinogalactan to peptidoglycan molecule.

Enoyl-[acyl-carrier-protein] reductase [NADH] (InhA) is involved in the biosynthesis of mycolic acids playing a key role in the fatty acid elongation process.

Decaprenylphosphoryl-beta-D-ribose oxidase (Dpre1) is a component of the complex that catalyzes the formation of decaprenyl-phospho-arabinose (DPA), a key precursor required for the synthesis of cell-wall arabinans.

N-Acetylglucosamine-1-phosphate uridyltransferase (GLMU) is an enzyme that catalyzes the final two steps in the biosynthesis of UDP-GlcNAc and is essential for the synthesis of the lipid A of lipopolysaccharide (LPS).

Shikimate kinase (AroK) is necessary for the synthesis of the common precursor of aromatic amino acids and secondary metabolites by phosphorylation of the 3-hydroxyl group of shikimic acid using ATP. The amino acids biosynthesis is also a very attractive direction for drug design, particularly as it is represented with a broad family of enzymes.

Beta-lactamase responsible for resistance to beta-lactam antibiotics.

Isocitrate lyase is in charge of the first step of the glyoxylate shunt, namely the synthesis of C4 dicarboxylic acids from C2 compounds.

ATPases (ClpC, ClpX) – chaperon proteins that induce degradation of proteins by bacterial proteasome. Play an essential role in bacterial proteostasis.

Mtb regulatory Serine Proteases (ClpP1P2) – serine proteases that form bacterial proteasomes and are responsible for protein homeostasis and degradation in bacteria.

Peptidyl tRNA hydrolase (Pth) is an essential Mtb-specific RNA hydrolase that participte in bacteria homeostasis and protein synthesis control.

The vHTS procedure assumes the application of grid-based representation of the binding site and designation of constraints based on literature data (mutation experiments) or ligand-protein interaction map with subsequent partial or full matching. Besides the rating of compounds by empirical docking score, the result conformations were also visually checked for the reliability of the binding mode. At the same time, the number of reference compounds against selected targets is relatively small and possesses a low degree of diversity inside each target-based reference set.

The example of the binding pose of Z1603549506 in the Shikimate kinase binding site.

Predicted binding conformation in ribbon and shape protein representation of Z1313265676 to RmlC reductase.

Examples of predicted hits Z111780734 (left) and Z189105794 binding conformation in the active site of InhA enzyme.

The dysregulation of Wnt/β-catenin pathway is often a key reason for cancer and other serious diseases. Wnt signaling is a highly conserved pathway that controls all steps of embryo development^1-6 and tissue maintenance in mature organisms^8-11. Alzheimer's, multiple sclerosis (MS), polycystic kidney disease (PKD), type 2 diabetes, arthritis, colon cancer, lung and breast cancer are some examples of Wnt pathway dysfunction. Recent evidence suggests that activation of Wnt pathway may prevent osteogenesis^17,30,31 and neurodegenerative disorders.^18,32

To create the ultimate Wnt/β-catenin signaling modulators library, we focused on the essential and clinically relevant protein targets. We used all available structural information and activity data for the following targets: CK1α, GSK3α/β, frizzled (Fzd) receptors and Lrp co-receptors (Lrp5/6), zinc and ring finger proteins (Znrf3, Rnf43), DKK1, Dvl-Axin, β-Catenin/TCF4, β-Catenin/Bcl9 and Bcl9/PYGO/His3 interaction inhibitions and Tankyrase to assemble the library.

The library of potential Wnt signaling modulators has been pre-plated for the most convenient and quick access. Using our Wnt library you receive multiple benefits, allowing you to save on time and costs in hit exploration and follow-up optimization:

• Hit confirmation support: Analogs search and hit samples resupply from same batch with strict QC
• Straightforward & affordable synthesis of hit follow-up libraries by parallel chemistry
• Medicinal chemistry support enhanced with on-site broad ADME/T panel

Figure 1. Canonical Wnt/Fz signaling cascade (adopted from Future Med. Chem. (2015), 7(18), 2485–2505) showing specific target-biased focused libraries available from Enamine.

Whereas numerous publications describe both identification and clinical development of Wnt inhibitors as potential anti-tumor therapies,^19,20 limited number of reports deal with the therapeutic utility of Wnt activators and/or potentiators.²¹ These classes of small molecule modulators usually inhibit kinases implicated in stability of β-catenin, namely GSK3β and CK1. For example, CHIR-99021 (CT99021) is a GSK-3α/βinhibitor with IC₅₀ of 10 nM/6.7 nM and > 500-fold selectivity for GSK-3 versus its closest homologs CDC2 and ERK2, as well as other protein kinases. Administration of CHIR-99021 significantly augmented hematopoietic repopulation in recipient mice transplanted with mouse or human hematopoietic stem cells (HSCs).²² This approach induces Wnt target gene expression via affecting the scaffolding proteins APC and Axin. However, GSK3 is involved in numerous signaling events in the cell and may lead to a mechanism-based toxicity. PF-670462 is a potent (IC₅₀ = 7.7 ± 2.2 nM) and selective (>30-fold with respect to 42 additional kinases) inhibitor of CK1ε that phosphorylates Dvl protein in the Wnt signaling.²³

There is the ongoing need for inhibitors, activators and potentiators (i.e. agents that mediate and/or enhance the effect of endogenous Wnt ligands) of Wnt signaling with novel mechanism of action. We combined our novel chemistries and structure-based approach to assemble a library of ca. 10,000 lead-like compounds able to tackle various aspects of Wnt/Fz signaling. Specific targets representing Wnt/Fz cascade include sFRP1, Dkk1, DyrK1A/B and interfaces of β-catenin with several transcription factors including Tcf-4 and Bcl-9. With this library we allow using either target-based, reporter-based or phenotypic assay(s) to test these compounds. Identified hits are expected to be tractable (i.e. feasible SAR and lead generation) and suitable as entry points into developing agents for the treatment of tumors, bone disease, Cardiovasular (CV) /metabolic, gastrointestinal (GI), and CNS/neurodegenerative disorders.

Specific targets and nodules of Fz/Wnt signaling cascade used for the library design:

Wnt/sFRP-1 interaction inhibitors. sFRP-1 family of proteins includes endogenous molecules that bind both Wnt ligands and Fz receptors. Inhibition of sFRP-1 may result in vicarious activation of Wnt signaling. In addition, WAY 316606 is a sFRP-1 inhibitor (Ki = 80 nM) that was described to increase total bone area in a murine calvarial organ culture assay at <1 nM concentration²⁴. It was also reported to be an antiosteoporotic agent²⁵. WAY-262611 is a Wnt pathway agonist that increases bone formation rate with EC₅₀ of 0.63 uM in TCF-Luciferase assay. WAY-262611 exhibits good pharmacokinetic properties and a dose-dependent increase in the trabecular bone formation rate in ovariectomized rats following oral administration²⁶. Specific attention needs to be paid to monitoring specific organ toxicity associated with a long-term perturbation of Wnt pathway including neoplasia(s),^27,28 although recent data suggest otherwise²⁹. This selection is represented by ca 1200 compounds.

Figure 2. Multifunctional molecular interactions of sFRP1/2 and their implications in pathological events. A) sFRP1/2 sequester Wnts, thereby antagonizing Wnt/β-catenin signaling; B) Frzb (sFRP3) and Crescent favor the diffusion of Wnts, enhancing their signaling range; C) sFRP1/2 cind to Fz receptors activating intracellular signaling; D) sFRP1 binds RANKL preventing its interaction with RANK receptor; E) sFRP2 binds to and enhances the procollagen proteinase activity of BMP1/Tollpid-like metalloproteinases accelerating the processing of pro-collagen; F) Sizzled binds to and inhibits the Chordinase activity of BMP1/Tolloid-like metalloproteinase; unproteced Chordin binds and sequesters BMPs (from Tohoku J. Exp. Med. 2010, 221, 11-17).

Figure 3. A) Pharmacophore model based on WAY 316606; orange – aromatic/heteroaromatic groups, green-hydrophobic pharmacophore (halogen, CF₃, CHF₂, Me, Et, iPr etc.), blue-positively charged moiety (⁺NH₃Alk, ⁺NH₂Alk₂, ⁺NHAlk₃ etc.); B) Superposition of WAY-316606 and our pharmacophore construct used in (sub) library design

DKK-1 Inhibitors. We used both docking and ligand-based approaches to search for potential DKK-1 binders. Specifically, the DKK1/LRP6 protein-protein interaction (PPI) interface was utilized to select matching ligands from the entire Enamine Screening Collection. The analogs of NCI8642 and other reported DKK-1/LRP6 interaction inhibitors were used for ligand-based search and pharmacophore modeling. In the docking calculation, we focused on topological and charge distribution features of the critical Dkk-1 loop that binds to LRP6. We reasoned that blocking/destabilizing this loop feature of the DKK-1/LRP6 PPI may yield modulators and/or activators of Wnt signaling while maintaining ‘normal’ Wnt-Fz-LRP6 signaling.

Figure 4. Binding interface between LRP6 protein and representative ligand from the focused subset.

We used known DKK-1 inhibitor WAY 262611 and its 3D pharmacophore in the ligand-based approach to identify 3k+ lead-like compounds as potential binders.

Figure 5. A) Pharmacophore model based on WAY-262611; orange – aromatic/heteroaromatic groups, blue – positively charged moiety (⁺NH₃Alk, ⁺NH₂Alk₂, ⁺NHAlk₃ etc.), magenta – aliphatic ring; B) Superposition of WAY-262611 and our pharmacophore construct used in (sub) library design.

Dvl-Axin Inhibitors. Dvl-Axin interaction is mediated by a specific PDZ-domain. Since structural information on this PPI interface is not available, we built a homology model using AlphaFold2 and numerous SBio data on PDZ domains. The resulting model was used to run docking calculation of the entire Screening Collection and identify the most promising ‘hits’. As a result, ~2500 final Dvl-biased compounds were added to the library. A representative example of putative Dvl binder is shown below.

Figure 6. A representative example of Dvl/PDZ-domain binding ligand from our focused set.

β-Catenin/TCF4 Protein-Protein Interaction. To model the interaction interface between β-catenin and Tcf-4 we used reported structural data (ACS Chem. Biol. 2014, 9, 193−201) to identify critical contacts including intriguing protonated pyridine (cation) – π-interaction between docked ligand and Arg474/515 tweezers. The topology of the hydrophobic pocket and H-bond of the aromatic NH₂ group with protein’s Lys508 residue allowed construction of a respective pharmacophore model (Figure 7). After docking calculation and scoring cut off, a focused selection of ca. 1800 small molecules was assembled into β-Catenin/TCF4 sublibrary.

Figure 7. Binding interface between β-catenin and representative ligand from the focused subset.

β-Catenin/Bcl9 Protein-Protein Interaction Interface. We focused on the available structural data of the reported helix-helix interaction between the two biological molecules. Specifically, we investigated the key H-bonds between H358, R359 (Bcl9) and D162, D164 (β-catenin) (Figure 8). Following docking and binding poses prioritization, we picked molecules that secured hydrogen bonding with D162/D164 and exhibited a significant π-stacking/hydrophobic interaction area with β-catenin to result in the 1779-compound library.

Figure 8. A) β-catenin/BCL9 helix-helix interaction interface featuring key hydrogen bonds; B) binding of β-catenin and a representative ligand from the focused subset.

Bcl9/PYGO/Histone 3 Interaction Interface. The other option to block Wnt/Fz signaling is disruption of PYGO/Histone 3 interaction (Figure 9A). Notably, it has been shown that small molecules could bind to the PHD domain of PYGO that engages the methylated Histone 3 residue, which leads to disruption of β-catenin/Bcl9 interaction. After docking calculation and scoring prioritization, 318 small molecules were selected as potential PPI inhibitors of this complex. The representative example of a ligand mediating multiple hydrophilic interactions, p-stacking with Phe366 and hydrogen bonds, is shown in Figure 9B.

Figure 9. A) Interaction interface of BCL9/PYGO/Histone 3; B) Representative molecule bound to the key amino acids of H3K4me2.

What Makes this Chemical Compound Library Unique and Valuable?

Most commercial compound libraries cover broad sectors of biology such as GPCRs, kinases, or antibacterials that may include hundreds to thousands of individual proteins. Here we provide targeted libraries, subsets from the Enamine REAL Space, which are designed around common features seen when grouping proteins by their binding properties. These AI-enabled libraries were built around the hundred most promising and clinically relevant drug targets grouped into 10 families.

Development Strategy

As a first step, the human proteome was encoded into binding likelihood vectors through MatchMaker, a neural network model trained on a vast collection of biochemical datasets, to predict interacting protein-ligand pairs, otherwise known as Drug-Target Interactions (DTI). The entire Enamine REAL Space library was then scanned for putative interactions within the proteome, and the resulting pairwise matrix (compound vs protein) of 2.8 quadrillion interactions was aggregated at the synthon level. The emerging space is rich in chemical and biological information. It can be navigated to identify chemo-proteomic clusters: small, focused target sets predicted to bind with structurally and chemically similar compounds.

• Targeted libraries are designed starting from a list of seed targets relevant to biological pathways of interest and expanded to relevant chemo-proteomic clusters.
• Compounds were ranked against targets in each cluster according to their MatchMaker DTI prediction score.
• The final library set was selected to maximize diversity in drug-like, synthesizable chemical space, through the application of a range of classical medicinal chemistry, DMSO stability and synthesizability filters.
• 15 000 new compounds were synthesized to deliver highly focused AI-driven libraries