Collaborative SAR Modeling and Prospective In Vitro Validation of Oxidative Stress Activation in Human HepG2 Cells.

Oxidative stress is the consequence of an abnormal increase of reactive oxygen species (ROS). ROS are generated mainly during the metabolism in both normal and pathological conditions as well as from exposure to xenobiotics. Xenobiotics can, on the one hand, disrupt molecular machinery involved in redox processes and, on the other hand, reduce the effectiveness of the antioxidant activity. Such dysregulation may lead to oxidative damage when combined with oxidative stress overpassing the cell capacity to detoxify ROS. In this work, a green fluorescent protein (GFP)-tagged nuclear factor erythroid 2-related factor 2 (NRF2)-regulated sulfiredoxin reporter (Srxn1-GFP) was used to measure the antioxidant response of HepG2 cells to a large series of drug and drug-like compounds (2230 compounds). These compounds were then classified as positive or negative depending on cellular response and distributed among different modeling groups to establish structure-activity relationship (SAR) models. A selection of models was used to prospectively predict oxidative stress induced by a new set of compounds subsequently experimentally tested to validate the model predictions. Altogether, this exercise exemplifies the different challenges of developing SAR models of a phenotypic cellular readout, model combination, chemical space selection, and results interpretation.

Large-Scale Modeling of Sparse Protein Kinase Activity Data.

Protein kinases are a protein family that plays an important role in several complex diseases such as cancer and cardiovascular and immunological diseases. Protein kinases have conserved ATP binding sites, which when targeted can lead to similar activities of inhibitors against different kinases. This can be exploited to create multitarget drugs. On the other hand, selectivity (lack of similar activities) is desirable in order to avoid toxicity issues. There is a vast amount of protein kinase activity data in the public domain, which can be used in many different ways. Multitask machine learning models are expected to excel for these kinds of data sets because they can learn from implicit correlations between tasks (in this case activities against a variety of kinases). However, multitask modeling of sparse data poses two major challenges: (i) creating a balanced train-test split without data leakage and (ii) handling missing data. In this work, we construct a protein kinase benchmark set composed of two balanced splits without data leakage, using random and dissimilarity-driven cluster-based mechanisms, respectively. This data set can be used for benchmarking and developing protein kinase activity prediction models. Overall, the performance on the dissimilarity-driven cluster-based split is lower than on random split-based sets for all models, indicating poor generalizability of models. Nevertheless, we show that multitask deep learning models, on this very sparse data set, outperform single-task deep learning and tree-based models. Finally, we demonstrate that data imputation does not improve the performance of (multitask) models on this benchmark set.

Enhancing Hit Discovery in Virtual Screening through Absolute Protein-Ligand Binding Free-Energy Calculations.

In the hit identification stage of drug discovery, a diverse chemical space needs to be explored to identify initial hits. Contrary to empirical scoring functions, absolute protein-ligand binding free-energy perturbation (ABFEP) provides a theoretically more rigorous and accurate description of protein-ligand binding thermodynamics and could, in principle, greatly improve the hit rates in virtual screening. In this work, we describe an implementation of an accurate and reliable ABFEP method in FEP+. We validated the ABFEP method on eight congeneric compound series binding to eight protein receptors including both neutral and charged ligands. For ligands with net charges, the alchemical ion approach is adopted to avoid artifacts in electrostatic potential energy calculations. The calculated binding free energies correlate with experimental results with a weighted average of R2 = 0.55 for the entire dataset. We also observe an overall root-mean-square error (RMSE) of 1.1 kcal/mol after shifting the zero-point of the simulation data to match the average experimental values. Through ABFEP calculations using apo versus holo protein structures, we demonstrated that the protein conformational and protonation state changes between the apo and holo proteins are the main physical factors contributing to the protein reorganization free energy manifested by the overestimation of raw ABFEP calculated binding free energies using the holo structures of the proteins. Furthermore, we performed ABFEP calculations in three virtual screening applications for hit enrichment. ABFEP greatly improves the hit rates as compared to docking scores or other methods like metadynamics. The good performance of ABFEP in rank ordering compounds demonstrated in this work confirms it as a useful tool to improve the hit rates in virtual screening, thus facilitating hit discovery.

Multitask machine learning models for predicting lipophilicity (logP) in the SAMPL7 challenge.

Accurate prediction of lipophilicity-logP-based on molecular structures is a well-established field. Predictions of logP are often used to drive forward drug discovery projects. Driven by the SAMPL7 challenge, in this manuscript we describe the steps that were taken to construct a novel machine learning model that can predict and generalize well. This model is based on the recently described Directed-Message Passing Neural Networks (D-MPNNs). Further enhancements included: both the inclusion of additional datasets from ChEMBL (RMSE improvement of 0.03), and the addition of helper tasks (RMSE improvement of 0.04). To the best of our knowledge, the concept of adding predictions from other models (Simulations Plus logP and logD@pH7.4, respectively) as helper tasks is novel and could be applied in a broader context. The final model that we constructed and used to participate in the challenge ranked 2/17 ranked submissions with an RMSE of 0.66, and an MAE of 0.48 (submission: Chemprop). On other datasets the model also works well, especially retrospectively applied to the SAMPL6 challenge where it would have ranked number one out of all submissions (RMSE of 0.35). Despite the fact that our model works well, we conclude with suggestions that are expected to improve the model even further.

Chemical genetics strategy to profile kinase target engagement reveals role of FES in neutrophil phagocytosis.

Chemical tools to monitor drug-target engagement of endogenously expressed protein kinases are highly desirable for preclinical target validation in drug discovery. Here, we describe a chemical genetics strategy to selectively study target engagement of endogenous kinases. By substituting a serine residue into cysteine at the DFG-1 position in the ATP-binding pocket, we sensitize the non-receptor tyrosine kinase FES towards covalent labeling by a complementary fluorescent chemical probe. This mutation is introduced in the endogenous FES gene of HL-60 cells using CRISPR/Cas9 gene editing. Leveraging the temporal and acute control offered by our strategy, we show that FES activity is dispensable for differentiation of HL-60 cells towards macrophages. Instead, FES plays a key role in neutrophil phagocytosis via SYK kinase activation. This chemical genetics strategy holds promise as a target validation method for kinases.

Accurate Prediction of GPCR Ligand Binding Affinity with Free Energy Perturbation.

The computational prediction of relative binding free energies is a crucial goal for drug discovery, and G protein-coupled receptors (GPCRs) are arguably the most important drug target class. However, they present increased complexity to model compared to soluble globular proteins. Despite breakthroughs, experimental X-ray crystal and cryo-EM structures are challenging to attain, meaning computational models of the receptor and ligand binding mode are sometimes necessary. This leads to uncertainty in understanding ligand-protein binding induced changes such as, water positioning and displacement, side chain positioning, hydrogen bond networks, and the overall structure of the hydration shell around the ligand and protein. In other words, the very elements that define structure activity relationships (SARs) and are crucial for accurate binding free energy calculations are typically more uncertain for GPCRs. In this work we use free energy perturbation (FEP) to predict the relative binding free energies for ligands of two different GPCRs. We pinpoint the key aspects for success such as the important role of key water molecules, amino acid ionization states, and the benefit of equilibration with specific ligands. Initial calculations following typical FEP setup and execution protocols delivered no correlation with experiment, but we show how results are improved in a logical and systematic way. This approach gave, in the best cases, a coefficient of determination (R2) compared with experiment in the range of 0.6-0.9 and mean unsigned errors compared to experiment of 0.6-0.7 kcal/mol. We anticipate that our findings will be applicable to other difficult-to-model protein ligand data sets and be of wide interest for the community to continue improving FE binding energy predictions.

Successive Statistical and Structure-Based Modeling to Identify Chemically Novel Kinase Inhibitors.

Kinases are frequently studied in the context of anticancer drugs. Their involvement in cell responses, such as proliferation, differentiation, and apoptosis, makes them interesting subjects in multitarget drug design. In this study, a workflow is presented that models the bioactivity spectra for two panels of kinases: (1) inhibition of RET, BRAF, SRC, and S6K, while avoiding inhibition of MKNK1, TTK, ERK8, PDK1, and PAK3, and (2) inhibition of AURKA, PAK1, FGFR1, and LKB1, while avoiding inhibition of PAK3, TAK1, and PIK3CA. Both statistical and structure-based models were included, which were thoroughly benchmarked and optimized. A virtual screening was performed to test the workflow for one of the main targets, RET kinase. This resulted in 5 novel and chemically dissimilar RET inhibitors with remaining RET activity of <60% (at a concentration of 10 µM) and similarities with known RET inhibitors from 0.18 to 0.29 (Tanimoto, ECFP6). The four more potent inhibitors were assessed in a concentration range and proved to be modestly active with a pIC50 value of 5.1 for the most active compound. The experimental validation of inhibitors for RET strongly indicates that the multitarget workflow is able to detect novel inhibitors for kinases, and hence, this workflow can potentially be applied in polypharmacology modeling. We conclude that this approach can identify new chemical matter for existing targets. Moreover, this workflow can easily be applied to other targets as well.

Synthesis and Pharmacological Evaluation of Triazolopyrimidinone Derivatives as Noncompetitive, Intracellular Antagonists for CC Chemokine Receptors 2 and 5.

CC chemokine receptors 2 (CCR2) and 5 (CCR5) are involved in many inflammatory diseases; however, most CCR2 and CCR5 clinical candidates have been unsuccessful. (Pre)clinical evidence suggests that dual CCR2/CCR5 inhibition might be more effective in the treatment of such multifactorial diseases. In this regard, the highly conserved intracellular binding site in chemokine receptors provides a new avenue for the design of multitarget ligands. In this study, we synthesized and evaluated the biological activity of a series of triazolopyrimidinone derivatives in CCR2 and CCR5. Radioligand binding assays first showed that they bind to the intracellular site of CCR2, and in combination with functional assays on CCR5, we explored structure-affinity/activity relationships in both receptors. Although most compounds were CCR2-selective, 39 and 43 inhibited ß-arrestin recruitment in CCR5 with high potency. Moreover, these compounds displayed an insurmountable mechanism of inhibition in both receptors, which holds promise for improved efficacy in inflammatory diseases.

Development of Covalent Ligands for G Protein-Coupled Receptors: A Case for the Human Adenosine A3 Receptor.

The development of covalent ligands for G protein-coupled receptors (GPCRs) is not a trivial process. Here, we report a streamlined workflow thereto from synthesis to validation, exemplified by the discovery of a covalent antagonist for the human adenosine A3 receptor (hA3AR). Based on the 1 H,3 H-pyrido[2,1- f]purine-2,4-dione scaffold, a series of ligands bearing a fluorosulfonyl warhead and a varying linker was synthesized. This series was subjected to an affinity screen, revealing compound 17b as the most potent antagonist. In addition, a nonreactive methylsulfonyl derivative 19 was developed as a reversible control compound. A series of assays, comprising time-dependent affinity determination, washout experiments, and [35S]GTPγS binding assays, then validated 17b as the covalent antagonist. A combined in silico hA3AR-homology model and site-directed mutagenesis study was performed to demonstrate that amino acid residue Y2657.36 was the unique anchor point of the covalent interaction. This workflow might be applied to other GPCRs to guide the discovery of covalent ligands.

Comprehensive structure-activity-relationship of azaindoles as highly potent FLT3 inhibitors.

Acute myeloid leukemia (AML) is characterized by fast progression and low survival rates, in which Fms-like tyrosine kinase 3 (FLT3) receptor mutations have been identified as a driver mutation in cancer progression in a subgroup of AML patients. Clinical trials have shown emergence of drug resistant mutants, emphasizing the ongoing need for new chemical matter to enable the treatment of this disease. Here, we present the discovery and topological structure-activity relationship (SAR) study of analogs of isoquinolinesulfonamide H-89, a well-known PKA inhibitor, as FLT3 inhibitors. Surprisingly, we found that the SAR was not consistent with the observed binding mode of H-89 in PKA. Matched molecular pair analysis resulted in the identification of highly active sub-nanomolar azaindoles as novel FLT3-inhibitors. Structure based modelling using the FLT3 crystal structure suggested an alternative, flipped binding orientation of the new inhibitors.

Drug Discovery Maps, a Machine Learning Model That Visualizes and Predicts Kinome-Inhibitor Interaction Landscapes.

The interpretation of high-dimensional structure-activity data sets in drug discovery to predict ligand-protein interaction landscapes is a challenging task. Here we present Drug Discovery Maps (DDM), a machine learning model that maps the activity profile of compounds across an entire protein family, as illustrated here for the kinase family. DDM is based on the t-distributed stochastic neighbor embedding (t-SNE) algorithm to generate a visualization of molecular and biological similarity. DDM maps chemical and target space and predicts the activities of novel kinase inhibitors across the kinome. The model was validated using independent data sets and in a prospective experimental setting, where DDM predicted new inhibitors for FMS-like tyrosine kinase 3 (FLT3), a therapeutic target for the treatment of acute myeloid leukemia. Compounds were resynthesized, yielding highly potent, cellularly active FLT3 inhibitors. Biochemical assays confirmed most of the predicted off-targets. DDM is further unique in that it is completely open-source and available as a ready-to-use executable to facilitate broad and easy adoption.

Pyrrolone Derivatives as Intracellular Allosteric Modulators for Chemokine Receptors: Selective and Dual-Targeting Inhibitors of CC Chemokine Receptors 1 and 2.

The recent crystal structures of CC chemokine receptors 2 and 9 (CCR2 and CCR9) have provided structural evidence for an allosteric, intracellular binding site. The high conservation of residues involved in this site suggests its presence in most chemokine receptors, including the close homologue CCR1. By using [3H]CCR2-RA-[ R], a high-affinity, CCR2 intracellular ligand, we report an intracellular binding site in CCR1, where this radioligand also binds with high affinity. In addition, we report the synthesis and biological characterization of a series of pyrrolone derivatives for CCR1 and CCR2, which allowed us to identify several high-affinity intracellular ligands, including selective and potential multitarget antagonists. Evaluation of selected compounds in a functional [35S]GTPγS assay revealed that they act as inverse agonists in CCR1, providing a new manner of pharmacological modulation. Thus, this intracellular binding site enables the design of selective and multitarget inhibitors as a novel therapeutic approach.

Intracellular Receptor Modulation: Novel Approach to Target GPCRs.

Recent crystal structures of multiple G protein-coupled receptors (GPCRs) have revealed a highly conserved intracellular pocket that can be used to modulate these receptors from the inside. This novel intracellular site partially overlaps with the G protein and ß-arrestin binding site, providing a new manner of pharmacological intervention. Here we provide an update of the architecture and function of the intracellular region of GPCRs, until now portrayed as the signaling domain. We review the available evidence on the presence of intracellular binding sites among chemokine receptors and other class A GPCRs, as well as different strategies to target it, including small molecules, pepducins, and nanobodies. Finally, the potential advantages of intracellular (allosteric) ligands over orthosteric ligands are also discussed.

Kinetics of human cannabinoid 1 (CB1) receptor antagonists: Structure-kinetics relationships (SKR) and implications for insurmountable antagonism.

While equilibrium binding affinities and in vitro functional antagonism of CB1 receptor antagonists have been studied in detail, little is known on the kinetics of their receptor interaction. In this study, we therefore conducted kinetic assays for nine 1-(4,5-diarylthiophene-2-carbonyl)-4-phenylpiperidine-4-carboxamide derivatives and included the CB1 antagonist rimonabant as a comparison. For this we newly developed a dual-point competition association assay with [3H]CP55940 as the radioligand. This assay yielded Kinetic Rate Index (KRI) values from which structure-kinetics relationships (SKR) of hCB1 receptor antagonists could be established. The fast dissociating antagonist 6 had a similar receptor residence time (RT) as rimonabant, i.e. 19 and 14 min, respectively, while the slowest dissociating antagonist (9) had a very long RT of 2222 min, i.e. pseudo-irreversible dissociation kinetics. In functional assays, 9 displayed insurmountable antagonism, while the effects of the shortest RT antagonist 6 and rimonabant were surmountable. Taken together, this study shows that hCB1 receptor antagonists can have very divergent RTs, which are not correlated to their equilibrium affinities. Furthermore, their RTs appear to define their mode of functional antagonism, i.e. surmountable vs. insurmountable. Finally, based on the recently resolved hCB1 receptor crystal structure, we propose that the differences in RT can be explained by a different binding mode of antagonist 9 from short RT antagonists that is able to displace unfavorable water molecules. Taken together, these findings are of importance for future design and evaluation of potent and safe hCB1 receptor antagonists.

Structure-Affinity Relationships and Structure-Kinetic Relationships of 1,2-Diarylimidazol-4-carboxamide Derivatives as Human Cannabinoid 1 Receptor Antagonists.

We report on the synthesis and biological evaluation of a series of 1,2-diarylimidazol-4-carboxamide derivatives developed as CB1 receptor antagonists. These were evaluated in a radioligand displacement binding assay, a [35S]GTPγS binding assay, and in a competition association assay that enables the relatively fast kinetic screening of multiple compounds. The compounds show high affinities and a diverse range of kinetic profiles at the CB1 receptor and their structure-kinetic relationships (SKRs) were established. Using the recently resolved hCB1 receptor crystal structures, we also performed a modeling study that sheds light on the crucial interactions for both the affinity and dissociation kinetics of this family of ligands. We provide evidence that, next to affinity, additional knowledge of binding kinetics is useful for selecting new hCB1 receptor antagonists in the early phases of drug discovery.

Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set.

The increase of publicly available bioactivity data in recent years has fueled and catalyzed research in chemogenomics, data mining, and modeling approaches. As a direct result, over the past few years a multitude of different methods have been reported and evaluated, such as target fishing, nearest neighbor similarity-based methods, and Quantitative Structure Activity Relationship (QSAR)-based protocols. However, such studies are typically conducted on different datasets, using different validation strategies, and different metrics. In this study, different methods were compared using one single standardized dataset obtained from ChEMBL, which is made available to the public, using standardized metrics (BEDROC and Matthews Correlation Coefficient). Specifically, the performance of Naïve Bayes, Random Forests, Support Vector Machines, Logistic Regression, and Deep Neural Networks was assessed using QSAR and proteochemometric (PCM) methods. All methods were validated using both a random split validation and a temporal validation, with the latter being a more realistic benchmark of expected prospective execution. Deep Neural Networks are the top performing classifiers, highlighting the added value of Deep Neural Networks over other more conventional methods. Moreover, the best method ('DNN_PCM') performed significantly better at almost one standard deviation higher than the mean performance. Furthermore, Multi-task and PCM implementations were shown to improve performance over single task Deep Neural Networks. Conversely, target prediction performed almost two standard deviations under the mean performance. Random Forests, Support Vector Machines, and Logistic Regression performed around mean performance. Finally, using an ensemble of DNNs, alongside additional tuning, enhanced the relative performance by another 27% (compared with unoptimized 'DNN_PCM'). Here, a standardized set to test and evaluate different machine learning algorithms in the context of multi-task learning is offered by providing the data and the protocols. Graphical Abstract .

Structure-Affinity Relationships and Structure-Kinetics Relationships of Pyrido[2,1-f]purine-2,4-dione Derivatives as Human Adenosine A3 Receptor Antagonists.

We expanded on a series of pyrido[2,1-f]purine-2,4-dione derivatives as human adenosine A3 receptor (hA3R) antagonists to determine their kinetic profiles and affinities. Many compounds showed high affinities and a diverse range of kinetic profiles. We found hA3R antagonists with very short residence time (RT) at the receptor (2.2 min for 5) and much longer RTs (e.g., 376 min for 27 or 391 min for 31). Two representative antagonists (5 and 27) were tested in [35S]GTPγS binding assays, and their RTs appeared correlated to their (in)surmountable antagonism. From a kon-koff-KD kinetic map, we divided the antagonists into three subgroups, providing a possible direction for the further development of hA3R antagonists. Additionally, we performed a computational modeling study that sheds light on the crucial receptor interactions, dictating the compounds' binding kinetics. Knowledge of target binding kinetics appears useful for developing and triaging new hA3R antagonists in the early phase of drug discovery.

Synthesis and evaluation of N-substituted 2-amino-4,5-diarylpyrimidines as selective adenosine A1 receptor antagonists.

We report the synthesis and biological evaluation of new 2-amino-4,5-diarylpyrimidines as selective antagonists at the adenosine A1 receptor. The scaffold they are based upon is a deaza variation of a previously reported collection of 3-amino-5,6-diaryl-1,2,4-triazines, members of which had a subnanomolar affinity but limited selectivity over the A2A subtype. Initially, similar structure-affinity relationships at the 5-aryl ring were established, and then emphasis was put on increasing selectivity at the hA1AR by introducing substituents on the N2-position, all the while maintaining a nanomolar affinity. Compound 3z, bearing a trans 4-hydroxycyclohexyl substituent, was identified as a potent (Ki(hA1AR) = 7.7 nM) and selective (Ki(hA2AAR) = 1389 nM) antagonist at the human adenosine A1 receptor. Computational docking was effected at the A1 and A2A subtypes, rationalizing the effect of the 4-hydroxycyclohexyl substituent on selectivity, in relation with the nature of the substituent on the 5-position of the pyrimidine.

A covalent antagonist for the human adenosine A2A receptor.

The structure of the human A2A adenosine receptor has been elucidated by X-ray crystallography with a high affinity non-xanthine antagonist, ZM241385, bound to it. This template molecule served as a starting point for the incorporation of reactive moieties that cause the ligand to covalently bind to the receptor. In particular, we incorporated a fluorosulfonyl moiety onto ZM241385, which yielded LUF7445 (4-((3-((7-amino-2-(furan-2-yl)-[1, 2, 4]triazolo[1,5-a][1, 3, 5]triazin-5-yl)amino)propyl)carbamoyl)benzene sulfonyl fluoride). In a radioligand binding assay, LUF7445 acted as a potent antagonist, with an apparent affinity for the hA2A receptor in the nanomolar range. Its apparent affinity increased with longer incubation time, suggesting an increasing level of covalent binding over time. An in silico A2A-structure-based docking model was used to study the binding mode of LUF7445. This led us to perform site-directed mutagenesis of the A2A receptor to probe and validate the target lysine amino acid K153 for covalent binding. Meanwhile, a functional assay combined with wash-out experiments was set up to investigate the efficacy of covalent binding of LUF7445. All these experiments led us to conclude LUF7445 is a valuable molecular tool for further investigating covalent interactions at this receptor. It may also serve as a prototype for a therapeutic approach in which a covalent antagonist may be needed to counteract prolonged and persistent presence of the endogenous ligand adenosine.

Relative Binding Free Energy Calculations Applied to Protein Homology Models.

A significant challenge and potential high-value application of computer-aided drug design is the accurate prediction of protein-ligand binding affinities. Free energy perturbation (FEP) using molecular dynamics (MD) sampling is among the most suitable approaches to achieve accurate binding free energy predictions, due to the rigorous statistical framework of the methodology, correct representation of the energetics, and thorough treatment of the important degrees of freedom in the system (including explicit waters). Recent advances in sampling methods and force fields coupled with vast increases in computational resources have made FEP a viable technology to drive hit-to-lead and lead optimization, allowing for more efficient cycles of medicinal chemistry and the possibility to explore much larger chemical spaces. However, previous FEP applications have focused on systems with high-resolution crystal structures of the target as starting points-something that is not always available in drug discovery projects. As such, the ability to apply FEP on homology models would greatly expand the domain of applicability of FEP in drug discovery. In this work we apply a particular implementation of FEP, called FEP+, on congeneric ligand series binding to four diverse targets: a kinase (Tyk2), an epigenetic bromodomain (BRD4), a transmembrane GPCR (A2A), and a protein-protein interaction interface (BCL-2 family protein MCL-1). We apply FEP+ using both crystal structures and homology models as starting points and find that the performance using homology models is generally on a par with the results when using crystal structures. The robustness of the calculations to structural variations in the input models can likely be attributed to the conformational sampling in the molecular dynamics simulations, which allows the modeled receptor to adapt to the "real" conformation for each ligand in the series. This work exemplifies the advantages of using all-atom simulation methods with full system flexibility and offers promise for the general application of FEP to homology models, although additional validation studies should be performed to further understand the limitations of the method and the scenarios where FEP will work best.