Discovery of a structural class of antibiotics with explainable deep learning.

The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis1-9. Deep learning approaches have aided in exploring chemical spaces1,10-15; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of 39,312 compounds and applied ensembles of graph neural networks to predict antibiotic activity and cytotoxicity for 12,076,365 compounds. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against Staphylococcus aureus were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant S. aureus (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.

Structural insight into allosteric modulation of protease-activated receptor 2.

Protease-activated receptors (PARs) are a family of G-protein-coupled receptors (GPCRs) that are irreversibly activated by proteolytic cleavage of the N terminus, which unmasks a tethered peptide ligand that binds and activates the transmembrane receptor domain, eliciting a cellular cascade in response to inflammatory signals and other stimuli. PARs are implicated in a wide range of diseases, such as cancer and inflammation. PARs have been the subject of major pharmaceutical research efforts but the discovery of small-molecule antagonists that effectively bind them has proved challenging. The only marketed drug targeting a PAR is vorapaxar, a selective antagonist of PAR1 used to prevent thrombosis. The structure of PAR1 in complex with vorapaxar has been reported previously. Despite sequence homology across the PAR isoforms, discovery of PAR2 antagonists has been less successful, although GB88 has been described as a weak antagonist. Here we report crystal structures of PAR2 in complex with two distinct antagonists and a blocking antibody. The antagonist AZ8838 binds in a fully occluded pocket near the extracellular surface. Functional and binding studies reveal that AZ8838 exhibits slow binding kinetics, which is an attractive feature for a PAR2 antagonist competing against a tethered ligand. Antagonist AZ3451 binds to a remote allosteric site outside the helical bundle. We propose that antagonist binding prevents structural rearrangements required for receptor activation and signalling. We also show that a blocking antibody antigen-binding fragment binds to the extracellular surface of PAR2, preventing access of the tethered ligand to the peptide-binding site. These structures provide a basis for the development of selective PAR2 antagonists for a range of therapeutic uses.

Discovery of cofactor-specific, bactericidal Mycobacterium tuberculosis InhA inhibitors using DNA-encoded library technology.

Millions of individuals are infected with and die from tuberculosis (TB) each year, and multidrug-resistant (MDR) strains of TB are increasingly prevalent. As such, there is an urgent need to identify novel drugs to treat TB infections. Current frontline therapies include the drug isoniazid, which inhibits the essential NADH-dependent enoyl-acyl-carrier protein (ACP) reductase, InhA. To inhibit InhA, isoniazid must be activated by the catalase-peroxidase KatG. Isoniazid resistance is linked primarily to mutations in the katG gene. Discovery of InhA inhibitors that do not require KatG activation is crucial to combat MDR TB. Multiple discovery efforts have been made against InhA in recent years. Until recently, despite achieving high potency against the enzyme, these efforts have been thwarted by lack of cellular activity. We describe here the use of DNA-encoded X-Chem (DEX) screening, combined with selection of appropriate physical properties, to identify multiple classes of InhA inhibitors with cell-based activity. The utilization of DEX screening allowed the interrogation of very large compound libraries (1011 unique small molecules) against multiple forms of the InhA enzyme in a multiplexed format. Comparison of the enriched library members across various screening conditions allowed the identification of cofactor-specific inhibitors of InhA that do not require activation by KatG, many of which had bactericidal activity in cell-based assays.

Discovery of a Potent BTK Inhibitor with a Novel Binding Mode by Using Parallel Selections with a DNA-Encoded Chemical Library.

We have identified and characterized novel potent inhibitors of Bruton's tyrosine kinase (BTK) from a single DNA-encoded library of over 110 million compounds by using multiple parallel selection conditions, including variation in target concentration and addition of known binders to provide competition information. Distinct binding profiles were observed by comparing enrichments of library building block combinations under these conditions; one enriched only at high concentrations of BTK and was competitive with ATP, and another enriched at both high and low concentrations of BTK and was not competitive with ATP. A compound representing the latter profile showed low nanomolar potency in biochemical and cellular BTK assays. Results from kinetic mechanism of action studies were consistent with the selection profiles. Analysis of the co-crystal structure of the most potent compound demonstrated a novel binding mode that revealed a new pocket in BTK. Our results demonstrate that profile-based selection strategies using DNA-encoded libraries form the basis of a new methodology to rapidly identify small molecule inhibitors with novel binding modes to clinically relevant targets.

The Development and Application of Biophysical Assays for Evaluating Ternary Complex Formation Induced by Proteolysis Targeting Chimeras (PROTACS).

E3 ligases and proteins targeted for degradation can be induced to form complexes by heterobifunctional molecules in a multi-step process. The kinetics and thermodynamics of the interactions involved contribute to efficiency of ubiquitination and resulting degradation of the protein. Biophysical techniques such as surface plasmon resonance (SPR), biolayer interferometry (BLI), and isothermal titration calorimetry (ITC) provide valuable information that can be used in the optimization of those interactions. Using two model systems, a biophysical assay tool kit for understanding the cooperativity of ternary complex formation and the impact of the 'hook effect' on binding kinetics was established. In one case, a proteolysis targeting chimera (PROTAC) molecule that induced ternary complex formation between Brd4BD2 and VHL was evaluated. The heterobifunctional molecule, MZ1, has nM affinities for both the Brd4BD2 protein (SPR KD = 1 nM, ITC KD = 4 nM) and the VHL complex (SPR KD = 29 nM, ITC KD = 66 nM). For this system, robust SPR, BLI, and ITC assays were developed that reproduced published results demonstrating the cooperativity of ternary complex formation. In the other case, a molecule that induced ternary complexes between a 46.0 kDa protein, PPM1D, and cereblon [CRBN (319-442)] was studied. The heterobifunctional molecule, BRD-5110, has an SPR KD = 1 nM for PPM1D but much weaker binding against the truncated CRBN (319-442) complex (SPR KD= ~ 3 µM). In that case, the binding for CRBN in SPR was not saturable, resulting in a "hook-effect". Throughput and reagent requirements for SPR, BLI, and ITC were evaluated, and general recommendations for their application to PROTAC projects were provided.

Imidazo[1,5-a]quinoxalines as irreversible BTK inhibitors for the treatment of rheumatoid arthritis.

Imidazo[1,5-a]quinoxalines were synthesized that function as irreversible Bruton's tyrosine kinase (BTK) inhibitors. The syntheses and SAR of this series of compounds are presented as well as the X-ray crystal structure of the lead compound 36 in complex with a gate-keeper variant of ITK enzyme. The lead compound showed good in vivo efficacy in preclinical RA models.

Discovery of Novel UDP-N-Acetylglucosamine Acyltransferase (LpxA) Inhibitors with Activity against Pseudomonas aeruginosa.

This study describes a novel series of UDP-N-acetylglucosamine acyltransferase (LpxA) inhibitors that was identified through affinity-mediated selection from a DNA-encoded compound library. The original hit was a selective inhibitor of Pseudomonas aeruginosa LpxA with no activity against Escherichia coli LpxA. The biochemical potency of the series was optimized through an X-ray crystallography-supported medicinal chemistry program, resulting in compounds with nanomolar activity against P. aeruginosa LpxA (best half-maximal inhibitory concentration (IC50) <5 nM) and cellular activity against P. aeruginosa (best minimal inhibitory concentration (MIC) of 4 µg/mL). Lack of activity against E. coli was maintained (IC50 > 20 µM and MIC > 128 µg/mL). The mode of action of analogues was confirmed through genetic analyses. As expected, compounds were active against multidrug-resistant isolates. Further optimization of pharmacokinetics is needed before efficacy studies in mouse infection models can be attempted. To our knowledge, this is the first reported LpxA inhibitor series with selective activity against P. aeruginosa.

Agonists and Antagonists of Protease-Activated Receptor 2 Discovered within a DNA-Encoded Chemical Library Using Mutational Stabilization of the Target.

The discovery of ligands via affinity-mediated selection of DNA-encoded chemical libraries is driven by the quality and concentration of the protein target. G-protein-coupled receptors (GPCRs) and other membrane-bound targets can be difficult to isolate in their functional state and at high concentrations, and therefore have been challenging for affinity-mediated selection. Here, we report a successful selection campaign against protease-activated receptor 2 (PAR2). Using a thermo-stabilized mutant of PAR2, we conducted affinity selection using our >100-billion-compound DNA-encoded library. We observed a number of putative ligands enriched upon selection, and subsequent cellular profiling revealed these ligands to comprise both agonists and antagonists. The agonist series shared structural similarity with known agonists. The antagonists were shown to bind in a novel allosteric binding site on the PAR2 protein. This report serves to demonstrate that cell-free affinity selection against GPCRs can be achieved with mutant stabilized protein targets.

Structure Based Design of Non-Natural Peptidic Macrocyclic Mcl-1 Inhibitors.

Mcl-1 is a pro-apoptotic BH3 protein family member similar to Bcl-2 and Bcl-xL. Overexpression of Mcl-1 is often seen in various tumors and allows cancer cells to evade apoptosis. Here we report the discovery and optimization of a series of non-natural peptide Mcl-1 inhibitors. Screening of DNA-encoded libraries resulted in hit compound 1, a 1.5 µM Mcl-1 inhibitor. A subsequent crystal structure demonstrated that compound 1 bound to Mcl-1 in a ß-turn conformation, such that the two ends of the peptide were close together. This proximity allowed for the linking of the two ends of the peptide to form a macrocycle. Macrocyclization resulted in an approximately 10-fold improvement in binding potency. Further exploration of a key hydrophobic interaction with Mcl-1 protein and also with the moiety that engages Arg256 led to additional potency improvements. The use of protein-ligand crystal structures and binding kinetics contributed to the design and understanding of the potency gains. Optimized compound 26 is a <3 nM Mcl-1 inhibitor, while inhibiting Bcl-2 at only 5 µM and Bcl-xL at >99 µM, and induces cleaved caspase-3 in MV4-11 cells with an IC50 of 3 µM after 6 h.

Correction to "Structure Based Design of Non-Natural Peptidic Macrocyclic Mcl-1 Inhibitors".

[This corrects the article DOI: 10.1021/acsmedchemlett.6b00464.].

Isoform-Selective ATAD2 Chemical Probe with Novel Chemical Structure and Unusual Mode of Action.

ATAD2 (ANCCA) is an epigenetic regulator and transcriptional cofactor, whose overexpression has been linked to the progress of various cancer types. Here, we report a DNA-encoded library screen leading to the discovery of BAY-850, a potent and isoform selective inhibitor that specifically induces ATAD2 bromodomain dimerization and prevents interactions with acetylated histones in vitro, as well as with chromatin in cells. These features qualify BAY-850 as a chemical probe to explore ATAD2 biology.

Discovery of Clinical Candidate 1-{[(2S,3S,4S)-3-Ethyl-4-fluoro-5-oxopyrrolidin-2-yl]methoxy}-7-methoxyisoquinoline-6-carboxamide (PF-06650833), a Potent, Selective Inhibitor of Interleukin-1 Receptor Associated Kinase 4 (IRAK4), by Fragment-Based Drug Design.

Through fragment-based drug design focused on engaging the active site of IRAK4 and leveraging three-dimensional topology in a ligand-efficient manner, a micromolar hit identified from a screen of a Pfizer fragment library was optimized to afford IRAK4 inhibitors with nanomolar potency in cellular assays. The medicinal chemistry effort featured the judicious placement of lipophilicity, informed by co-crystal structures with IRAK4 and optimization of ADME properties to deliver clinical candidate PF-06650833 (compound 40). This compound displays a 5-unit increase in lipophilic efficiency from the fragment hit, excellent kinase selectivity, and pharmacokinetic properties suitable for oral administration.

Overview of Recent Progress in Protein-Expression Technologies for Small-Molecule Screening.

Production of novel soluble and membrane-localized protein targets for functional and affinity-based screening has often been limited by the inability of traditional protein-expression systems to generate recombinant proteins that have properties similar to those of their endogenous counterparts. Such targets have often been labeled as challenging. Although biological validation of these challenging targets for specific disease areas may be strong, discovery of small-molecule modulators can be greatly delayed or completely halted due to target-expression issues. In this article, the limitations of traditional protein-expression systems will be discussed along with new systems designed to overcome these challenges. Recent work in this field has focused on two major areas for both soluble and membrane targets: construct-design strategies to improve expression levels and new hosts that can carry out the posttranslational modifications necessary for proper target folding and function. Another area of active research has been on the reconstitution of solubilized membrane targets for both structural analysis and screening. Finally, the potential impact of these new systems on the output of small-molecule screening campaigns will be discussed.

Structure guided development of novel thymidine mimetics targeting Pseudomonas aeruginosa thymidylate kinase: from hit to lead generation.

Thymidylate kinase (TMK) is a potential chemotherapeutic target because it is directly involved in the synthesis of an essential component, thymidine triphosphate, in DNA replication. All reported TMK inhibitors are thymidine analogues, which might retard their development as potent therapeutics due to cell permeability and off-target activity against human TMK. A small molecule hit (1, IC(50) = 58 µM), which has reasonable inhibition potency against Pseudomonas aeruginosa TMK (PaTMK), was identified by the analysis of the binding mode of thymidine or TP(5)A in a PaTMK homology model. This hit (1) was cocrystallized with PaTMK, and several potent PaTMK inhibitors (leads, 46, 47, 48, and 56, IC(50) = 100-200 nM) were synthesized using computer-aided design approaches including virtual synthesis/screening, which was used to guide the design of inhibitors. The binding mode of the optimized leads in PaTMK overlaps with that of other bacterial TMKs but not with human TMK, which shares few common features with the bacterial enzymes. Therefore, the optimized TMK inhibitors described here should be useful for the development of antibacterial agents targeting TMK without undesired off-target effects. In addition, an inhibition mechanism associated with the LID loop, which mimics the process of phosphate transfer from ATP to dTMP, was proposed based on X-ray cocrystal structures, homology models, and structure-activity relationship results.

Fragment-based screen against HIV protease.

We have employed a fragment-based screen against wild-type (NL4-3) HIV protease (PR) using the Active Sight fragment library and X-ray crystallography. The experiments reveal two new binding sites for small molecules. PR was co-crystallized with fragments, or crystals were soaked in fragment solutions, using five crystal forms, and 378 data sets were collected to 2.3-1.3 A resolution. Fragment binding induces a distinct conformation and specific crystal form of TL-3 inhibited PR during co-crystallization. One fragment, 2-methylcyclohexanol, binds in the 'exo site' adjacent to the Gly(16)Gly(17)Gln(18)loop where the amide of Gly(17)is a specific hydrogen bond donor, and hydrophobic contacts occur with the side chains of Lys(14)and Leu(63). Another fragment, indole-6-carboxylic acid, binds on the 'outside/top of the flap' via hydrophobic contacts with Trp(42), Pro(44), Met(46), and Lys(55), a hydrogen bond with Val(56), and a salt-bridge with Arg(57). 2-acetyl-benzothiophene also binds at this site. This study is the first fragment-based crystallographic screen against HIV PR, and the first time that fragments were screened against an inhibitor-bound drug target to search for compounds that both bind to novel sites and stabilize the inhibited conformation of the target.