Structure-based, deep-learning models for protein-ligand binding affinity prediction

The launch of AlphaFold series has brought deep-learning techniques into the molecular structural science. As another crucial problem, structure-based prediction of protein-ligand binding affinity urgently calls for advanced computational techniques. Is deep learning ready to decode this problem? Here we review mainstream structure-based, deep-learning approaches for this problem, focusing on molecular representations, learning architectures and model interpretability. A model taxonomy has been generated. To compensate for the lack of valid comparisons among those models, we realized and evaluated representatives from a uniform basis, with the advantages and shortcomings discussed. This review will potentially benefit structure-based drug discovery and related areas.

ChemFlow_py: a flexible toolkit for docking and rescoring

The design of accurate virtual screening tools is an open challenge in drug discovery. Several structure-based methods have been developed at different levels of approximation. Among them, molecular docking is an established technique with high efficiency, but typically low accuracy. Moreover, docking performances are known to be target-dependent, which makes the choice of the docking program and corresponding scoring function critical when approaching a new protein target. To compare the performances of different docking protocols, we developed ChemFlow_py, an automated tool to perform docking and rescoring. Using four protein systems extracted from DUD-E with 100 known active compounds and 3000 decoys per target, we compared the performances of several rescoring strategies including consensus scoring. We found that the average docking results can be improved by consensus ranking, which emphasizes the relevance of consensus scoring when little or no chemical information is available for a given target. ChemFlow_py is a free toolkit to optimize the performances of virtual high-throughput screening (vHTS). The software is publicly available at https://github.com/IFMlab/ChemFlow_py .

Merging cultures and disciplines to create a drug discovery ecosystem at Virginia commonwealth university: Medicinal chemistry, structural biology, molecular and behavioral pharmacology and computational chemistry

The Department of Medicinal Chemistry, together with the Institute for Structural Biology, Drug Discovery and Development, at Virginia Commonwealth University (VCU) has evolved, organically with quite a bit of bootstrapping, into a unique drug discovery ecosystem in response to the environment and culture of the university and the wider research enterprise. Each faculty member that joined the department and/or institute added a layer of expertise, technology and most importantly, innovation, that fertilized numerous collaborations within the University and with outside partners. Despite moderate institutional support with respect to a typical drug discovery enterprise, the VCU drug discovery ecosystem has built and maintained an impressive array of facilities and instrumentation for drug synthesis, drug characterization, biomolecular structural analysis and biophysical analysis, and pharmacological studies. Altogether, this ecosystem has had major impacts on numerous therapeutic areas, such as neurology, psychiatry, drugs of abuse, cancer, sickle cell disease, coagulopathy, inflammation, aging disorders and others. Novel tools and strategies for drug discovery, design and development have been developed at VCU in the last five decades; e.g., fundamental rational structure-activity relationship (SAR)-based drug design, structure-based drug design, orthosteric and allosteric drug design, design of multi-functional agents towards polypharmacy outcomes, principles on designing glycosaminoglycans as drugs, and computational tools and algorithms for quantitative SAR (QSAR) and understanding the roles of water and the hydrophobic effect.

Structure-based discovery and in vitro validation of inhibitors of chloride intracellular channel 4 protein

The use of computer-aided methods have continued to propel accelerated drug discovery across various disease models, interestingly allowing the specific inhibition of pathogenic targets. Chloride Intracellular Channel Protein 4 (CLIC4) is a novel class of intracellular ion channel highly implicated in tumor and vascular biology. It regulates cell proliferation, apoptosis and angiogenesis; and is involved in multiple pathologic signaling pathways. Absence of specific inhibitors however impedes its advancement to translational research. Here, we integrate structural bioinformatics and experimental research approaches for the discovery and validation of small-molecule inhibitors of CLIC4. High-affinity allosteric binders were identified from a library of 1615 Food and Drug Administration (FDA)-approved drugs via a high-performance computing-powered blind-docking approach, resulting in the selection of amphotericin B and rapamycin. NMR assays confirmed the binding and conformational disruptive effects of both drugs while they also reversed stress-induced membrane translocation of CLIC4 and inhibited endothelial cell migration. Structural and dynamics simulation studies further revealed that the inhibitory mechanisms of these compounds were hinged on the allosteric modulation of the catalytic glutathione (GSH)-like site loop and the extended catalytic β loop which may elicit interference with the catalytic activities of CLIC4. Structure-based insights from this study provide the basis for the selective targeting of CLIC4 to treat the associated pathologies.

CSAlign and CSAlign-Dock: Structure alignment of ligands considering full flexibility and application to protein-ligand docking

Structure prediction of protein-ligand complexes, called protein-ligand docking, is a critical computational technique that can be used to understand the underlying principle behind the protein functions at the atomic level and to design new molecules regulating the functions. Protein-ligand docking methods have been employed in structure-based drug discovery for hit discovery and lead optimization. One of the important technical challenges in protein-ligand docking is to account for protein conformational changes induced by ligand binding. A small change such as a single side-chain rotation upon ligand binding can hinder accurate docking. Here we report an increase in docking performance achieved by structure alignment to known complex structures. First, a fully flexible compound-to-compound alignment method CSAlign is developed by global optimization of a shape score. Next, the alignment method is combined with a docking algorithm to dock a new ligand to a target protein when a reference protein-ligand complex structure is available. This alignment-based docking method, called CSAlign-Dock, showed superior performance to ab initio docking methods in cross-docking benchmark tests. Both CSAlign and CSAlign-Dock are freely available as a web server at https://galaxy.seoklab.org/csalign.

Crystal structures of the catalytic domain of human PARP15 in complex with small molecule inhibitors

PARP15, or ARTD7, is an enzyme carrying out mono-ADP-ribosylation and regulating activities of a range of cellular proteins. This enzyme belongs to the family of the poly(ADP-ribose) polymerases (PARPs), which comprises of proteins with various potential disease indications. Due to their involvement in a number of cellular processes and important role in DNA repair and regulation, PARPs have been considered attractive therapeutic targets over the past few years. The pursuit of small molecule PARP inhibitors has resulted in several FDA approved drugs for multiple cancers so far. As the use of PARP inhibitors as drug scaffolds is actively explored recently, there is increasing interest in the design of selective inhibitors based on the structural features of the PARP proteins. Here, we solved high-resolution crystal structures of the human PARP15 catalytic domain in complex with three marketed drugs of PARP inhibitors, which includes compounds 3-AB, iniparib and niraparib. The structures reported here contribute to our understanding of the ligand binding modes and structural features in the PARP15 catalytic domain, which can be employed to guide the rational design of selective inhibitors of PARPs.

Fragment- and structure-based drug discovery for developing therapeutic agents targeting the DNA Damage Response

Cancer will directly affect the lives of over one-third of the population. The DNA Damage Response (DDR) is an intricate system involving damage recognition, cell cycle regulation, DNA repair, and ultimately cell fate determination, playing a central role in cancer etiology and therapy. Two primary therapeutic approaches involving DDR targeting include: combinatorial treatments employing anticancer genotoxic agents; and synthetic lethality, exploiting a sporadic DDR defect as a mechanism for cancer-specific therapy. Whereas, many DDR proteins have proven "undruggable", Fragment- and Structure-Based Drug Discovery (FBDD, SBDD) have advanced therapeutic agent identification and development. FBDD has led to 4 (with ∼50 more drugs under preclinical and clinical development), while SBDD is estimated to have contributed to the development of >200, FDA-approved medicines. Protein X-ray crystallography-based fragment library screening, especially for elusive or "undruggable" targets, allows for simultaneous generation of hits plus details of protein-ligand interactions and binding sites (orthosteric or allosteric) that inform chemical tractability, downstream biology, and intellectual property. Using a novel high-throughput crystallography-based fragment library screening platform, we screened five diverse proteins, yielding hit rates of ∼2-8% and crystal structures from ∼1.8 to 3.2 Å. We consider current FBDD/SBDD methods and some exemplary results of efforts to design inhibitors against the DDR nucleases meiotic recombination 11 (MRE11, a.k.a., MRE11A), apurinic/apyrimidinic endonuclease 1 (APE1, a.k.a., APEX1), and flap endonuclease 1 (FEN1).

Structure-based discovery of potent and selective small-molecule inhibitors targeting signal transducer and activator of transcription 3 (STAT3)

STAT3 has been validated as an attractive anticancer target due to its important roles in cancer initiation and progression. However, discovery of potent and selective STAT3 small-molecule inhibitors with druglike properties is still challenging. In this study, two series of substituted 2-phenylquinolines and 2-arylimidazo[1,2-a]pyridines were designed through structure-based drug discovery approach by condensing the privileged structures of STX-119 and SH4-54. Our study has resulted in the discovery of a number of highly potent and selective STAT3 inhibitors, exemplified by compound 39 with the privileged structure of 2-phenylimidazo[1,2-a]pyridine, which selectively inhibits phosphorylation of STAT3 and suppresses subsequent signaling pathway. Moreover, 39 inhibits cell growth, migration and invasion of human triple negative breast cancer (TNBC) cells lines. Consistently, it achieves significant and dose-dependent tumor growth inhibition in both cell line-derived and patient-derived xenograft tumor models in mice. These results clearly indicate that 39 is a highly potent and selective STAT3 inhibitor.

The critical role of QM/MM X-ray refinement and accurate tautomer/protomer determination in structure-based drug design

Conventional protein:ligand crystallographic refinement uses stereochemistry restraints coupled with a rudimentary energy functional to ensure the correct geometry of the model of the macromolecule—along with any bound ligand(s)—within the context of the experimental, X-ray density. These methods generally lack explicit terms for electrostatics, polarization, dispersion, hydrogen bonds, and other key interactions, and instead they use pre-determined parameters (e.g. bond lengths, angles, and torsions) to drive structural refinement. In order to address this deficiency and obtain a more complete and ultimately more accurate structure, we have developed an automated approach for macromolecular refinement based on a two layer, QM/MM (ONIOM) scheme as implemented within our DivCon Discovery Suite and "plugged in" to two mainstream crystallographic packages: PHENIX and BUSTER. This implementation is able to use one or more region layer(s), which is(are) characterized using linear-scaling, semi-empirical quantum mechanics, followed by a system layer which includes the balance of the model and which is described using a molecular mechanics functional. In this work, we applied our Phenix/DivCon refinement method—coupled with our XModeScore method for experimental tautomer/protomer state determination—to the characterization of structure sets relevant to structure-based drug design (SBDD). We then use these newly refined structures to show the impact of QM/MM X-ray refined structure on our understanding of function by exploring the influence of these improved structures on protein:ligand binding affinity prediction (and we likewise show how we use post-refinement scoring outliers to inform subsequent X-ray crystallographic efforts). Through this endeavor, we demonstrate a computational chemistry ↔ structural biology (X-ray crystallography) "feedback loop" which has utility in industrial and academic pharmaceutical research as well as other allied fields.

Structural basis for GPCR signaling by small polar versus large lipid metabolites-discovery of non-metabolite ligands

Key metabolites act through specific G protein-coupled receptors (GPCRs) as extracellular signals of fuel availability and metabolic stress. Here, we focus on the succinate receptor SUCNR1/GPR91 and the long chain fatty acid receptor FFAR1/GPR40, for which 3D structural information is available. Like other small polar acidic metabolites, succinate is excreted from the cell by transporter proteins to bind to an extracellular, solvent-exposed pocket in SUCNR1. Non-metabolite pharmacological tool compounds are currently being designed based on the structure of the SUCNR1 binding pocket. In FFAR1, differently signaling lipid mimetics bind in two distinct membrane-exposed sites corresponding to each of the lipid bilayer leaflets. Conceivably endogenous lipid ligands gain access to these sites by way of the membrane and probably occupy both sites under physiological circumstances. Design of polar agonists for a dynamic, solvent-exposed pocket in FFAR1 underlines the possibility of structure-based approaches for development of novel tool compounds even in lipid sensing metabolite GPCRs.

Protein Allostery in Rational Drug Design

This chapter focuses on protein kinases that transfer the phosphate group of ATP to the hydroxyl group of a substrate protein. Five hundred eighteen human protein kinases are classified into serine/threonine kinases and tyrosine kinases and individually or synergistically transduce physiologic stimuli into cell to promote cell proliferation or apoptosis, etc. Protein kinases are identified as drug targets because dysfunction of kinases leads to severe diseases such as cancers and autoimmune diseases. A large number of the crystal structures of the protein kinase inhibitor complex are available in Protein Data Bank and facilitated the drug discovery targeting protein kinases. The protein kinase inhibitors are classified into categories, Type-I, Type-II, Type-III, Type-IV, and Type-V, and as a separate class, covalent-type inhibitors. In any type, a protein kinase inhibitor bound to the allosteric region is advantageous in terms of selectivity compared to the traditional ATP-competitive one. In the following sections, the successful and promising examples of the partially or fully allosteric protein kinase inhibitors are illustrated in the following pages.

Alkynamide phthalazinones as a new class of TbrPDEB1 inhibitors

Several 3',5'-cyclic nucleotide phosphodiesterases (PDEs) have been validated as good drug targets for a large variety of diseases. Trypanosoma brucei PDEB1 (TbrPDEB1) has been designated as a promising drug target for the treatment of human African trypanosomiasis. Recently, the first class of selective nanomolar TbrPDEB1 inhibitors was obtained by targeting the parasite specific P-pocket. However, these biphenyl-substituted tetrahydrophthalazinone-based inhibitors did not show potent cellular activity against Trypanosoma brucei (T. brucei) parasites, leaving room for further optimization. Herein, we report the discovery of a new class of potent TbrPDEB1 inhibitors that display improved activities against T. brucei parasites. Exploring different linkers between the reported tetrahydrophthalazinone core scaffold and the amide tail group resulted in the discovery of alkynamide phthalazinones as new TbrPDEB1 inhibitors, which exhibit submicromolar activities versus T. brucei parasites and no cytotoxicity to human MRC-5 cells. Elucidation of the crystal structure of alkynamide 8b (NPD-048) bound to the catalytic domain of TbrPDEB1 shows a bidentate interaction with the key-residue Gln874 and good directionality towards the P-pocket. Incubation of trypanosomes with alkynamide 8b results in an increase of intracellular cAMP, validating a PDE-mediated effect in vitro and providing a new interesting compound series for further studies towards selective TbrPDEB1 inhibitors with potent phenotypic activity.

Aminobenzothiazole derivatives stabilize the thermolabile p53 cancer mutant Y220C and show anticancer activity in p53-Y220C cell lines

Many cancers have the tumor suppressor p53 inactivated by mutation, making reactivation of mutant p53 with small molecules a promising strategy for the development of novel anticancer therapeutics. The oncogenic p53 mutation Y220C, which accounts for approximately 100,000 cancer cases per year, creates an extended surface crevice in the DNA-binding domain, which destabilizes p53 and causes denaturation and aggregation. Here, we describe the structure-guided design of a novel class of small-molecule Y220C stabilizers and the challenging synthetic routes developed in the process. The synthesized chemical probe MB710, an aminobenzothiazole derivative, binds tightly to the Y220C pocket and stabilizes p53-Y220C in vitro. MB725, an ethylamide analogue of MB710, induced selective viability reduction in several p53-Y220C cancer cell lines while being well tolerated in control cell lines. Reduction of viability correlated with increased and selective transcription of p53 target genes such as BTG2, p21, PUMA, FAS, TNF, and TNFRSF10B, which promote apoptosis and cell cycle arrest, suggesting compound-mediated transcriptional activation of the Y220C mutant. Our data provide a framework for the development of a class of potent, non-toxic compounds for reactivating the Y220C mutant in anticancer therapy.

Integrating Experimental and In Silico HTS in the Discovery of Inhibitors of Protein-Nucleic Acid Interactions

Discovery of novel tool compounds and drug leads against a range of unorthodox protein targets has pushed both experimental screening methodologies as well as the field of structure-based design to the limit in recent years. Increasingly, it has been recognized that some of the most desirable targets for the development of small-molecule effectors are actually protein-protein and protein-nucleic acid interactions. There are numerous nontrivial challenges to pursuing small-molecule lead compounds directed toward PPIs and PNIs: relatively shallow cavities, large surface areas that are natively complexed to macromolecules, complex patterns of interstitial waters, a paucity of "hot spots," large conformational changes upon ligand binding, etc. Although there have been some notable successes targeting PPIs in the last decade, there has been distinctly less success in the realm of targeting PNIs. This chapter focuses on an approach, successfully applied by our group to address the challenge of gaining traction on the PPI target RAD52, which is a protein that binds both single-stranded and double-stranded DNA, and is an anticancer target for certain types of cancer. There are many approaches to tackling the difficult problems of finding effective small molecules that disrupt PPIs and PNIs, but the methods presented here offer a series of elegant solutions, which integrate experimental HTS, biophysical methods, docking, and molecular dynamics in a powerful way. Additionally, the structural knowledge gained from these studies provides a means for rationally understanding what features lead to ligand affinity in these fascinating and highly unorthodox pockets.

Predictive Power of Different Types of Experimental Restraints in Small Molecule Docking: A Review

Incorporating experimental restraints is a powerful method of increasing accuracy in computational protein small molecule docking simulations. Different algorithms integrate distinct forms of biochemical data during the docking and/or scoring stages. These so-called hybrid methods make use of receptor-based information such as nuclear magnetic resonance (NMR) restraints or small molecule-based information such as structure-activity relationships (SARs). A third class of methods directly interrogates contacts between the protein receptor and the small molecule. This work reviews the current state of using such restraints in docking simulations, evaluates their feasibility across broad systems, and identifies potential areas of algorithm development.

Current and Future Challenges in GPCR Drug Discovery

GPCRs play a pervasive physiological role and, in turn, are the leading target class for pharmaceuticals. Beginning with the determination of the structure of rhodopsin, and dramatically accelerating since the reporting of the first ligand-mediated GPCR X-ray structures, our understanding of the structural and functional characteristics of these proteins has grown dramatically. Deploying this now rapidly emerging information for drug discovery has already been extensively demonstrated through a watershed of studies appearing in numerous scientific reports. Included in these expositions are areas such as sites and characteristics of ligand to GPCR binding, protein activation, effector bias, allosteric mechanisms, dimerization, polypharmacology and others. Computational chemistry studies are demonstrating an increasing role in capitalizing on the structural studies to further advance our understanding of these proteins as well as to drive drug discovery. Such drug discovery activities range from the design of orthosteric site inhibitors through, for example, allosteric modulators, biased ligands, partial agonists and bitopic ligands. Herein, these topics are outlined through specific examples in the hopes of providing a glimpse of the state of the field.

From a novel HTS hit to potent, selective, and orally bioavailable KDM5 inhibitors

A high-throughput screening (HTS) of the Genentech/Roche library identified a novel, uncharged scaffold as a KDM5A inhibitor. Lacking insight into the binding mode, initial attempts to improve inhibitor potency failed to improve potency, and synthesis of analogs was further hampered by the presence of a C-C bond between the pyrrolidine and pyridine. Replacing this with a C-N bond significantly simplified synthesis, yielding pyrazole analog 35, of which we obtained a co-crystal structure with KDM5A. Using structure-based design approach, we identified 50 with improved biochemical, cell potency and reduced MW and lower lipophilicity (LogD) compared with the original hit. Furthermore, 50 showed lower clearance than 9 in mice. In combination with its remarkably low plasma protein binding (PPB) in mice (40%), oral dosing of 50 at 5mg/kg resulted in unbound C_max ∼2-fold of its cell potency (PC9 H3K4Me3 0.96μM), meeting our criteria for an in vivo tool compound from a new scaffold.

Discovery of TAK-659 an orally available investigational inhibitor of Spleen Tyrosine Kinase (SYK)

Spleen Tyrosine Kinase (SYK) is a non-receptor cytoplasmic tyrosine kinase that is primarily expressed in hematopoietic cells. SYK is a key mediator for a variety of inflammatory cells, including B cells, mast cells, macrophages and neutrophils and therefore, an attractive approach for treatment of both inflammatory diseases and oncology indications. Using in house co-crystal structure information, and structure-based drug design, we designed and optimized a novel series of heteroaromatic pyrrolidinone SYK inhibitors resulting in the selection of the development candidate TAK-659. TAK-659 is currently undergoing Phase I clinical trials for advanced solid tumor and lymphoma malignancies, a Phase Ib study in advanced solid tumors in combination with nivolumab, and PhIb/II trials for relapsed/refractory AML.

The Ascaris suum nicotinic receptor, ACR-16, as a drug target: Four novel negative allosteric modulators from virtual screening

Soil-transmitted helminth infections in humans and livestock cause significant debility, reduced productivity and economic losses globally. There are a limited number of effective anthelmintic drugs available for treating helminths infections, and their frequent use has led to the development of resistance in many parasite species. There is an urgent need for novel therapeutic drugs for treating these parasites. We have chosen the ACR-16 nicotinic acetylcholine receptor of Ascaris suum (Asu-ACR-16), as a drug target and have developed three-dimensional models of this transmembrane protein receptor to facilitate the search for new bioactive compounds. Using the human α7 nAChR chimeras and Torpedo marmorata nAChR for homology modeling, we defined orthosteric and allosteric binding sites on the Asu-ACR-16 receptor for virtual screening. We identified four ligands that bind to sites on Asu-ACR-16 and tested their activity using electrophysiological recording from Asu-ACR-16 receptors expressed in Xenopus oocytes. The four ligands were acetylcholine inhibitors (SB-277011-A, IC₅₀, 3.12 ± 1.29 μM; (+)-butaclamol Cl, IC₅₀, 9.85 ± 2.37 μM; fmoc-1, IC₅₀, 10.00 ± 1.38 μM; fmoc-2, IC₅₀, 16.67 ± 1.95 μM) that behaved like negative allosteric modulators. Our work illustrates a structure-based in silico screening method for seeking anthelmintic hits, which can then be tested electrophysiologically for further characterization.

•

Three-dimensional structural models of the Ascaris nicotinic (Asu-ACR-16) receptor made by homology modeling.

•

High affinity ligands selected by in silico screening.

•

Four ligands validated by electrophysiological studies as negative allosteric modulators.