MolScore: a scoring, evaluation and benchmarking framework for generative models in de novo drug design.

Generative models are undergoing rapid research and application to de novo drug design. To facilitate their application and evaluation, we present MolScore. MolScore already contains many drug-design-relevant scoring functions commonly used in benchmarks such as, molecular similarity, molecular docking, predictive models, synthesizability, and more. In addition, providing performance metrics to evaluate generative model performance based on the chemistry generated. With this unification of functionality, MolScore re-implements commonly used benchmarks in the field (such as GuacaMol, MOSES, and MolOpt). Moreover, new benchmarks can be created trivially. We demonstrate this by testing a chemical language model with reinforcement learning on three new tasks of increasing complexity related to the design of 5-HT2a ligands that utilise either molecular descriptors, 266 pre-trained QSAR models, or dual molecular docking. Lastly, MolScore can be integrated into an existing Python script with just three lines of code. This framework is a step towards unifying generative model application and evaluation as applied to drug design for both practitioners and researchers. The framework can be found on GitHub and downloaded directly from the Python Package Index.Scientific ContributionMolScore is an open-source platform to facilitate generative molecular design and evaluation thereof for application in drug design. This platform takes important steps towards unifying existing benchmarks, providing a platform to share new benchmarks, and improves customisation, flexibility and usability for practitioners over existing solutions.

Pharmacological Characterization and Radiolabeling of VUF15485, a High-Affinity Small-Molecule Agonist for the Atypical Chemokine Receptor ACKR3.

Atypical chemokine receptor 3 (ACKR3), formerly referred to as CXCR7, is considered to be an interesting drug target. In this study, we report on the synthesis, pharmacological characterization and radiolabeling of VUF15485, a new ACKR3 small-molecule agonist, that will serve as an important new tool to study this ß-arrestin-biased chemokine receptor. VUF15485 binds with nanomolar affinity (pIC50 = 8.3) to human ACKR3, as measured in [125I]CXCL12 competition binding experiments. Moreover, in a bioluminescence resonance energy transfer-based ß-arrestin2 recruitment assay VUF15485 acts as a potent ACKR3 agonist (pEC50 = 7.6) and shows a similar extent of receptor activation compared with CXCL12 when using a newly developed, fluorescence resonance energy transfer-based ACKR3 conformational sensor. Moreover, the ACKR3 agonist VUF15485, tested against a (atypical) chemokine receptor panel (agonist and antagonist mode), proves to be selective for ACKR3. VUF15485 labeled with tritium at one of its methoxy groups ([3H]VUF15485), binds ACKR3 saturably and with high affinity (K d = 8.2 nM). Additionally, [3H]VUF15485 shows rapid binding kinetics and consequently a short residence time (<2 minutes) for binding to ACKR3. The selectivity of [3H]VUF15485 for ACKR3, was confirmed by binding studies, whereupon CXCR3, CXCR4, and ACKR3 small-molecule ligands were competed for binding against the radiolabeled agonist. Interestingly, the chemokine ligands CXCL11 and CXCL12 are not able to displace the binding of [3H]VUF15485 to ACKR3. The radiolabeled VUF15485 was subsequently used to evaluate its binding pocket. Site-directed mutagenesis and docking studies using a recently solved cryo-EM structure propose that VUF15485 binds in the major and the minor binding pocket of ACKR3. SIGNIFICANCE STATEMENT: The atypical chemokine receptor atypical chemokine receptor 3 (ACKR3) is considered an interesting drug target in relation to cancer and multiple sclerosis. The study reports on new chemical biology tools for ACKR3, i.e., a new agonist that can also be radiolabeled and a new ACKR3 conformational sensor, that both can be used to directly study the interaction of ACKR3 ligands with the G protein-coupled receptor.

Small Molecule Fluorescent Ligands for the Atypical Chemokine Receptor 3 (ACKR3).

The atypical chemokine receptor 3 (ACKR3) is a receptor that induces cancer progression and metastasis in multiple cell types. Therefore, new chemical tools are required to study the role of ACKR3 in cancer and other diseases. In this study, fluorescent probes, based on a series of small molecule ACKR3 agonists, were synthesized. Three fluorescent probes, which showed specific binding to ACKR3 through a luminescence-based NanoBRET binding assay (pKd ranging from 6.8 to 7.8) are disclosed. Due to their high affinity at the ACKR3, we have shown their application in both competition binding experiments and confocal microscopy studies showing the cellular distribution of this receptor.

Alchemical Free Energy Calculations on Membrane-Associated Proteins.

Membrane proteins have diverse functions within cells and are well-established drug targets. The advances in membrane protein structural biology have revealed drug and lipid binding sites on membrane proteins, while computational methods such as molecular simulations can resolve the thermodynamic basis of these interactions. Particularly, alchemical free energy calculations have shown promise in the calculation of reliable and reproducible binding free energies of protein-ligand and protein-lipid complexes in membrane-associated systems. In this review, we present an overview of representative alchemical free energy studies on G-protein-coupled receptors, ion channels, transporters as well as protein-lipid interactions, with emphasis on best practices and critical aspects of running these simulations. Additionally, we analyze challenges and successes when running alchemical free energy calculations on membrane-associated proteins. Finally, we highlight the value of alchemical free energy calculations calculations in drug discovery and their applicability in the pharmaceutical industry.

Predicting the target landscape of kinase inhibitors using 3D convolutional neural networks.

Many therapies in clinical trials are based on single drug-single target relationships. To further extend this concept to multi-target approaches using multi-targeted drugs, we developed a machine learning pipeline to unravel the target landscape of kinase inhibitors. This pipeline, which we call 3D-KINEssence, uses a new type of protein fingerprints (3D FP) based on the structure of kinases generated through a 3D convolutional neural network (3D-CNN). These 3D-CNN kinase fingerprints were matched to molecular Morgan fingerprints to predict the targets of each respective kinase inhibitor based on available bioactivity data. The performance of the pipeline was evaluated on two test sets: a sparse drug-target set where each drug is matched in most cases to a single target and also on a densely-covered drug-target set where each drug is matched to most if not all targets. This latter set is more challenging to train, given its non-exclusive character. Our model's root-mean-square error (RMSE) based on the two datasets was 0.68 and 0.8, respectively. These results indicate that 3D FP can predict the target landscape of kinase inhibitors at around 0.8 log units of bioactivity. Our strategy can be utilized in proteochemometric or chemogenomic workflows by consolidating the target landscape of kinase inhibitors.

Integrating structure-based approaches in generative molecular design.

Generative molecular design for drug discovery and development has seen a recent resurgence promising to improve the efficiency of the design-make-test-analyse cycle; by computationally exploring much larger chemical spaces than traditional virtual screening techniques. However, most generative models thus far have only utilized small-molecule information to train and condition de novo molecule generators. Here, we instead focus on recent approaches that incorporate protein structure into de novo molecule optimization in an attempt to maximize the predicted on-target binding affinity of generated molecules. We summarize these structure integration principles into either distribution learning or goal-directed optimization and for each case whether the approach is protein structure-explicit or implicit with respect to the generative model. We discuss recent approaches in the context of this categorization and provide our perspective on the future direction of the field.

SAR exploration of the non-imidazole histamine H3 receptor ligand ZEL-H16 reveals potent inverse agonism.

Histamine H3 receptor (H3 R) agonists without an imidazole moiety remain very scarce. Of these, ZEL-H16 (1) has been reported previously as a high-affinity non-imidazole H3 R (partial) agonist. Our structure-activity relationship analysis using derivatives of 1 identified both basic moieties as key interaction motifs and the distance of these from the central core as a determinant for H3 R affinity. However, in spite of the reported H3 R (partial) agonism, in our hands, 1 acts as an inverse agonist for Gαi signaling in a CRE-luciferase reporter gene assay and using an H3 R conformational sensor. Inverse agonism was also observed for all of the synthesized derivatives of 1. Docking studies and molecular dynamics simulations suggest ionic interactions/hydrogen bonds to H3 R residues D1143.32 and E2065.46 as essential interaction points.

Augmented Hill-Climb increases reinforcement learning efficiency for language-based de novo molecule generation.

A plethora of AI-based techniques now exists to conduct de novo molecule generation that can devise molecules conditioned towards a particular endpoint in the context of drug design. One popular approach is using reinforcement learning to update a recurrent neural network or language-based de novo molecule generator. However, reinforcement learning can be inefficient, sometimes requiring up to 105 molecules to be sampled to optimize more complex objectives, which poses a limitation when using computationally expensive scoring functions like docking or computer-aided synthesis planning models. In this work, we propose a reinforcement learning strategy called Augmented Hill-Climb based on a simple, hypothesis-driven hybrid between REINVENT and Hill-Climb that improves sample-efficiency by addressing the limitations of both currently used strategies. We compare its ability to optimize several docking tasks with REINVENT and benchmark this strategy against other commonly used reinforcement learning strategies including REINFORCE, REINVENT (version 1 and 2), Hill-Climb and best agent reminder. We find that optimization ability is improved ~ 1.5-fold and sample-efficiency is improved ~ 45-fold compared to REINVENT while still delivering appealing chemistry as output. Diversity filters were used, and their parameters were tuned to overcome observed failure modes that take advantage of certain diversity filter configurations. We find that Augmented Hill-Climb outperforms the other reinforcement learning strategies used on six tasks, especially in the early stages of training or for more difficult objectives. Lastly, we show improved performance not only on recurrent neural networks but also on a reinforcement learning stabilized transformer architecture. Overall, we show that Augmented Hill-Climb improves sample-efficiency for language-based de novo molecule generation conditioning via reinforcement learning, compared to the current state-of-the-art. This makes more computationally expensive scoring functions, such as docking, more accessible on a relevant timescale.

Classification Model for the Second Extracellular Loop of Class A GPCRs.

The extracellular loop 2 (ECL2) is the longest and the most diverse loop among class A G protein-coupled receptors (GPCRs). It connects the transmembrane (TM) helices 4 and 5 and contains a highly conserved cysteine through which it is bridged with TM3. In this paper, experimental ECL2 structures were analyzed based on their sequences, shapes, and intramolecular contacts. To take into account the flexibility, we incorporated into our analyses information from the molecular dynamics trajectories available on the GPCRmd website. Despite the high sequence variability, shapes of the analyzed structures, defined by the backbone volume overlaps, can be clustered into seven main groups. Conformational differences within the clusters can be then identified by intramolecular interactions with other GPCR structural domains. Overall, our work provides a reorganization of the structural information of the ECL2 of class A GPCR subfamilies, highlighting differences and similarities on sequence and conformation levels.

Applications of Artificial Intelligence in Drug Design: Opportunities and Challenges.

Artificial intelligence (AI) has undergone rapid development in recent years and has been successfully applied to real-world problems such as drug design. In this chapter, we review recent applications of AI to problems in drug design including virtual screening, computer-aided synthesis planning, and de novo molecule generation, with a focus on the limitations of the application of AI therein and opportunities for improvement. Furthermore, we discuss the broader challenges imposed by AI in translating theoretical practice to real-world drug design; including quantifying prediction uncertainty and explaining model behavior.

Structure-Based Virtual Screening for Ligands of G Protein-Coupled Receptors: What Can Molecular Docking Do for You?

G protein-coupled receptors (GPCRs) constitute the largest family of membrane proteins in the human genome and are important therapeutic targets. During the last decade, the number of atomic-resolution structures of GPCRs has increased rapidly, providing insights into drug binding at the molecular level. These breakthroughs have created excitement regarding the potential of using structural information in ligand design and initiated a new era of rational drug discovery for GPCRs. The molecular docking method is now widely applied to model the three-dimensional structures of GPCR-ligand complexes and screen for chemical probes in large compound libraries. In this review article, we first summarize the current structural coverage of the GPCR superfamily and the understanding of receptor-ligand interactions at atomic resolution. We then present the general workflow of structure-based virtual screening and strategies to discover GPCR ligands in chemical libraries. We assess the state of the art of this research field by summarizing prospective applications of virtual screening based on experimental structures. Strategies to identify compounds with specific efficacy and selectivity profiles are discussed, illustrating the opportunities and limitations of the molecular docking method. Our overview shows that structure-based virtual screening can discover novel leads and will be essential in pursuing the next generation of GPCR drugs. SIGNIFICANCE STATEMENT: Extraordinary advances in the structural biology of G protein-coupled receptors have revealed the molecular details of ligand recognition by this large family of therapeutic targets, providing novel avenues for rational drug design. Structure-based docking is an efficient computational approach to identify novel chemical probes from large compound libraries, which has the potential to accelerate the development of drug candidates.

From structure to clinic: Design of a muscarinic M1 receptor agonist with potential to treatment of Alzheimer's disease.

Current therapies for Alzheimer's disease seek to correct for defective cholinergic transmission by preventing the breakdown of acetylcholine through inhibition of acetylcholinesterase, these however have limited clinical efficacy. An alternative approach is to directly activate cholinergic receptors responsible for learning and memory. The M1-muscarinic acetylcholine (M1) receptor is the target of choice but has been hampered by adverse effects. Here we aimed to design the drug properties needed for a well-tolerated M1-agonist with the potential to alleviate cognitive loss by taking a stepwise translational approach from atomic structure, cell/tissue-based assays, evaluation in preclinical species, clinical safety testing, and finally establishing activity in memory centers in humans. Through this approach, we rationally designed the optimal properties, including selectivity and partial agonism, into HTL9936-a potential candidate for the treatment of memory loss in Alzheimer's disease. More broadly, this demonstrates a strategy for targeting difficult GPCR targets from structure to clinic.

Discovery of Novel Allosteric Modulators Targeting an Extra-Helical Binding Site of GLP-1R Using Structure- and Ligand-Based Virtual Screening.

Allosteric modulators have emerged with many potential pharmacological advantages as they do not compete the binding of agonist or antagonist to the orthosteric sites but ultimately affect downstream signaling. To identify allosteric modulators targeting an extra-helical binding site of the glucagon-like peptide-1 receptor (GLP-1R) within the membrane environment, the following two computational approaches were applied: structure-based virtual screening with consideration of lipid contacts and ligand-based virtual screening with the maintenance of specific allosteric pocket residue interactions. Verified by radiolabeled ligand binding and cAMP accumulation experiments, two negative allosteric modulators and seven positive allosteric modulators were discovered using structure-based and ligand-based virtual screening methods, respectively. The computational approach presented here could possibly be used to discover allosteric modulators of other G protein-coupled receptors.

Comparison of structure- and ligand-based scoring functions for deep generative models: a GPCR case study.

Deep generative models have shown the ability to devise both valid and novel chemistry, which could significantly accelerate the identification of bioactive compounds. Many current models, however, use molecular descriptors or ligand-based predictive methods to guide molecule generation towards a desirable property space. This restricts their application to relatively data-rich targets, neglecting those where little data is available to sufficiently train a predictor. Moreover, ligand-based approaches often bias molecule generation towards previously established chemical space, thereby limiting their ability to identify truly novel chemotypes. In this work, we assess the ability of using molecular docking via Glide-a structure-based approach-as a scoring function to guide the deep generative model REINVENT and compare model performance and behaviour to a ligand-based scoring function. Additionally, we modify the previously published MOSES benchmarking dataset to remove any induced bias towards non-protonatable groups. We also propose a new metric to measure dataset diversity, which is less confounded by the distribution of heavy atom count than the commonly used internal diversity metric. With respect to the main findings, we found that when optimizing the docking score against DRD2, the model improves predicted ligand affinity beyond that of known DRD2 active molecules. In addition, generated molecules occupy complementary chemical and physicochemical space compared to the ligand-based approach, and novel physicochemical space compared to known DRD2 active molecules. Furthermore, the structure-based approach learns to generate molecules that satisfy crucial residue interactions, which is information only available when taking protein structure into account. Overall, this work demonstrates the advantage of using molecular docking to guide de novo molecule generation over ligand-based predictors with respect to predicted affinity, novelty, and the ability to identify key interactions between ligand and protein target. Practically, this approach has applications in early hit generation campaigns to enrich a virtual library towards a particular target, and also in novelty-focused projects, where de novo molecule generation either has no prior ligand knowledge available or should not be biased by it.

Controlling the selectivity of aminergic GPCR ligands from the extracellular vestibule.

In addition to the orthosteric binding pocket (OBP) of GPCRs, recent structural studies have revealed that there are several allosteric sites available for pharmacological intervention. The secondary binding pocket (SBP) of aminergic GPCRs is located in the extracellular vestibule of these receptors, and it has been suggested to be a potential selectivity pocket for bitopic ligands. Here, we applied a virtual screening protocol based on fragment docking to the SBP of the orthosteric ligand-receptor complex. This strategy was employed for a number of aminergic receptors. First, we designed dopamine D3 preferring bitopic compounds from a D2 selective orthosteric ligand. Next, we designed 5-HT2B selective bitopic compounds starting from the 5-HT1B preferring ergoline core of LSD. Comparing the serotonergic profiles of the new derivatives to that of LSD, we found that these derivatives became significantly biased towards the desired 5-HT2B receptor target. Finally, addressing the known limitations of H1 antihistamines, our protocol was successfully used to eliminate the well-known side effects related to the muscarinic M1 activity of amitriptyline while preserving H1 potency in some of the designed bitopic compounds. These applications highlight the usefulness of our new virtual screening protocol and offer a powerful strategy towards bitopic GPCR ligands with designed receptor profiles.

Vanishing white matter: Eukaryotic initiation factor 2B model and the impact of missense mutations.

BACKGROUND: Vanishing white matter (VWM) is a leukodystrophy, caused by recessive mutations in eukaryotic initiation factor 2B (eIF2B)-subunit genes (EIF2B1-EIF2B5); 80% are missense mutations. Clinical severity is highly variable, with a strong, unexplained genotype-phenotype correlation. MATERIALS AND METHODS: With information from a recent natural history study, we severity-graded 97 missense mutations. Using in silico modeling, we created a new human eIF2B model structure, onto which we mapped the missense mutations. Mutated residues were assessed for location in subunits, eIF2B complex, and functional domains, and for information on biochemical activity. RESULTS: Over 50% of mutations have (ultra-)severe phenotypic effects. About 60% affect the ε-subunit, containing the catalytic domain, mostly with (ultra-)severe effects. About 55% affect subunit cores, with variable clinical severity. About 36% affect subunit interfaces, mostly with severe effects. Very few mutations occur on the external eIf2B surface, perhaps because they have minor functional effects and are tolerated. One external surface mutation affects eIF2B-substrate interaction and is associated with ultra-severe phenotype. CONCLUSION: Mutations that lead to (ultra-)severe disease mostly affect amino acids with pivotal roles in complex formation and function of eIF2B. Therapies for VWM are emerging and reliable mutation-based phenotype prediction is required for propensity score matching for trials and in the future for individualized therapy decisions.

KLIFS: an overhaul after the first 5 years of supporting kinase research.

Kinases are a prime target of drug development efforts with >60 drug approvals in the past two decades. Due to the research into this protein family, a wealth of data has been accumulated that keeps on growing. KLIFS-Kinase-Ligand Interaction Fingerprints and Structures-is a structural database focusing on how kinase inhibitors interact with their targets. The aim of KLIFS is to support (structure-based) kinase research through the systematic collection, annotation, and processing of kinase structures. Now, 5 years after releasing the initial KLIFS website, the database has undergone a complete overhaul with a new website, new logo, and new functionalities. In this article, we start by looking back at how KLIFS has been used by the research community, followed by a description of the renewed KLIFS, and conclude with showcasing the functionalities of KLIFS. Major changes include the integration of approved drugs and inhibitors in clinical trials, extension of the coverage to atypical kinases, and a RESTful API for programmatic access. KLIFS is available at the new domain https://klifs.net.

Advanced fluorescence microscopy reveals disruption of dynamic CXCR4 dimerization by subpocket-specific inverse agonists.

Although class A G protein-coupled receptors (GPCRs) can function as monomers, many of them form dimers and oligomers, but the mechanisms and functional relevance of such oligomerization is ill understood. Here, we investigate this problem for the CXC chemokine receptor 4 (CXCR4), a GPCR that regulates immune and hematopoietic cell trafficking, and a major drug target in cancer therapy. We combine single-molecule microscopy and fluorescence fluctuation spectroscopy to investigate CXCR4 membrane organization in living cells at densities ranging from a few molecules to hundreds of molecules per square micrometer of the plasma membrane. We observe that CXCR4 forms dynamic, transient homodimers, and that the monomer-dimer equilibrium is governed by receptor density. CXCR4 inverse agonists that bind to the receptor minor pocket inhibit CXCR4 constitutive activity and abolish receptor dimerization. A mutation in the minor binding pocket reduced the dimer-disrupting ability of these ligands. In addition, mutating critical residues in the sixth transmembrane helix of CXCR4 markedly diminished both basal activity and dimerization, supporting the notion that CXCR4 basal activity is required for dimer formation. Together, these results link CXCR4 dimerization to its density and to its activity. They further suggest that inverse agonists binding to the minor pocket suppress both dimerization and constitutive activity and may represent a specific strategy to target CXCR4.