Protein Structure-Based Organic Chemistry-Driven Ligand Design from Ultralarge Chemical Spaces.

Ultralarge chemical spaces describing several billion compounds are revolutionizing hit identification in early drug discovery. Because of their size, such chemical spaces cannot be fully enumerated and require ad-hoc computational tools to navigate them and pick potentially interesting hits. We here propose a structure-based approach to ultralarge chemical space screening in which commercial chemical reagents are first docked to the target of interest and then directly connected according to organic chemistry and topological rules, to enumerate drug-like compounds under three-dimensional constraints of the target. When applied to bespoke chemical spaces of different sizes and chemical complexity targeting two receptors of pharmaceutical interest (estrogen ß receptor, dopamine D3 receptor), the computational method was able to quickly enumerate hits that were either known ligands (or very close analogs) of targeted receptors as well as chemically novel candidates that could be experimentally confirmed by in vitro binding assays. The proposed approach is generic, can be applied to any docking algorithm, and requires few computational resources to prioritize easily synthesizable hits from billion-sized chemical spaces.

Overlap of On-demand Ultra-large Combinatorial Spaces with On-the-shelf Drug-like Libraries.

On-demand combinatorial spaces are shifting paradigms in early drug discovery, by considerably increasing the searchable chemical space to several billions of compounds while securing their synthetic accessibility. We here systematically compared the on-the-shelf available drug-like chemical space (9 million compounds) to three on-demand ultra-large (ODUL) combinatorial fragment spaces (REAL, CHEMriya, GalaXi) covering 32 billion of readily accessible molecules. Surprisingly, only one space (REAL) intersects almost entirely the currently available drug-like space, suggesting that it is the only ODUL widely suitable for in-stock hit expansion. Of course, expanding a preliminary ODUL hit in the same chemical space is the best possible strategy to rapidly generate structure-activity relationships. All three spaces remain well suited to early hit finding initiatives since they all provide numerous unique scaffolds that are not described by on-the shelf collections.

Target-Focused Library Design by Pocket-Applied Computer Vision and Fragment Deep Generative Linking.

We here describe a computational approach (POEM: Pocket Oriented Elaboration of Molecules) to drive the generation of target-focused libraries while taking advantage of all publicly available structural information on protein-ligand complexes. A collection of 31 384 PDB-derived images with key shapes and pharmacophoric properties, describing fragment-bound microenvironments, is first aligned to the query target cavity by a computer vision method. The fragments of the most similar PDB subpockets are then directly positioned in the query cavity using the corresponding image transformation matrices. Lastly, suitable connectable atoms of oriented fragment pairs are linked by a deep generative model to yield fully connected molecules. POEM was applied to generate a library of 1.5 million potential cyclin-dependent kinase 8 inhibitors. By synthesizing and testing as few as 43 compounds, a few nanomolar inhibitors were quickly obtained with limited resources in just two iterative cycles.

Estimating the Similarity between Protein Pockets.

With the exponential increase in publicly available protein structures, the comparison of protein binding sites naturally emerged as a scientific topic to explain observations or generate hypotheses for ligand design, notably to predict ligand selectivity for on- and off-targets, explain polypharmacology, and design target-focused libraries. The current review summarizes the state-of-the-art computational methods applied to pocket detection and comparison as well as structural druggability estimates. The major strengths and weaknesses of current pocket descriptors, alignment methods, and similarity search algorithms are presented. Lastly, an exhaustive survey of both retrospective and prospective applications in diverse medicinal chemistry scenarios illustrates the capability of the existing methods and the hurdle that still needs to be overcome for more accurate predictions.

On the Frustration to Predict Binding Affinities from Protein-Ligand Structures with Deep Neural Networks.

Accurate prediction of binding affinities from protein-ligand atomic coordinates remains a major challenge in early stages of drug discovery. Using modular message passing graph neural networks describing both the ligand and the protein in their free and bound states, we unambiguously evidence that an explicit description of protein-ligand noncovalent interactions does not provide any advantage with respect to ligand or protein descriptors. Simple models, inferring binding affinities of test samples from that of the closest ligands or proteins in the training set, already exhibit good performances, suggesting that memorization largely dominates true learning in the deep neural networks. The current study suggests considering only noncovalent interactions while omitting their protein and ligand atomic environments. Removing all hidden biases probably requires much denser protein-ligand training matrices and a coordinated effort of the drug design community to solve the necessary protein-ligand structures.

High-Throughput Screening for Extracellular Inhibitors of the FLT3 Receptor Tyrosine Kinase Reveals Chemically Diverse and Druggable Negative Allosteric Modulators.

Inhibiting receptor tyrosine kinases is commonly achieved by two main strategies targeting either the intracellular kinase domain by low molecular weight compounds or the extracellular ligand-binding domain by monoclonal antibodies. Identifying small molecules able to inhibit RTKs at the extracellular level would be highly desirable to gain exquisite selectivity but is believed to be challenging owing to the size of RTK endogenous ligands (cytokines, growth factors) and the topology of RTK extracellular domains. We here report the high-throughput screening of the French Chemical Library (48K compounds) for extracellular inhibitors of the Fms-like tyrosine kinase 3 (FLT3) receptor tyrosine kinase, by a homogeneous time-resolved fluorescence competition assay. A total of 679 small molecular weight ligands (1.4%) were confirmed to strongly inhibit (>75%) the binding of the fluorescent labeled FLT3 ligand (FL cytokine) to FLT3 overexpressed in HEK-293 cells, at two different concentrations (5 and 20 µM). Concentration-response curves, obtained for 111 lead-like molecules, confirmed the unexpected tolerance of the FLT3 extracellular domain for low molecular weight druggable inhibitors exhibiting submicromolar potencies, chemical diversity, and promising pharmacokinetic properties. Further investigation of one hit confirmed inhibitory properties in dorsal root ganglia neurons and in a mouse model of neuropathic pain.

Targeting the Central Pocket of the Pseudomonas aeruginosa Lectin LecA.

Pseudomonas aeruginosa is an opportunistic ESKAPE pathogen that produces two lectins, LecA and LecB, as part of its large arsenal of virulence factors. Both carbohydrate-binding proteins are central to the initial and later persistent infection processes, i. e. bacterial adhesion and biofilm formation. The biofilm matrix is a major resistance determinant and protects the bacteria against external threats such as the host immune system or antibiotic treatment. Therefore, the development of drugs against the P. aeruginosa biofilm is of particular interest to restore efficacy of antimicrobials. Carbohydrate-based inhibitors for LecA and LecB were previously shown to efficiently reduce biofilm formations. Here, we report a new approach for inhibiting LecA with synthetic molecules bridging the established carbohydrate-binding site and a central cavity located between two LecA protomers of the lectin tetramer. Inspired by in silico design, we synthesized various galactosidic LecA inhibitors with aromatic moieties targeting this central pocket. These compounds reached low micromolar affinities, validated in different biophysical assays. Finally, X-ray diffraction analysis revealed the interactions of this compound class with LecA. This new mode of action paves the way to a novel route towards inhibition of P. aeruginosa biofilms.

Targeting undruggable carbohydrate recognition sites through focused fragment library design.

Carbohydrate-protein interactions are key for cell-cell and host-pathogen recognition and thus, emerged as viable therapeutic targets. However, their hydrophilic nature poses major limitations to the conventional development of drug-like inhibitors. To address this shortcoming, four fragment libraries were screened to identify metal-binding pharmacophores (MBPs) as novel scaffolds for inhibition of Ca2+-dependent carbohydrate-protein interactions. Here, we show the effect of MBPs on the clinically relevant lectins DC-SIGN, Langerin, LecA and LecB. Detailed structural and biochemical investigations revealed the specificity of MBPs for different Ca2+-dependent lectins. Exploring the structure-activity relationships of several fragments uncovered the functional groups in the MBPs suitable for modification to further improve lectin binding and selectivity. Selected inhibitors bound efficiently to DC-SIGN-expressing cells. Altogether, the discovery of MBPs as a promising class of Ca2+-dependent lectin inhibitors creates a foundation for fragment-based ligand design for future drug discovery campaigns.

Unexpected similarity between HIV-1 reverse transcriptase and tumor necrosis factor binding sites revealed by computer vision.

Rationalizing the identification of hidden similarities across the repertoire of druggable protein cavities remains a major hurdle to a true proteome-wide structure-based discovery of novel drug candidates. We recently described a new computational approach (ProCare), inspired by numerical image processing, to identify local similarities in fragment-based subpockets. During the validation of the method, we unexpectedly identified a possible similarity in the binding pockets of two unrelated targets, human tumor necrosis factor alpha (TNF-α) and HIV-1 reverse transcriptase (HIV-1 RT). Microscale thermophoresis experiments confirmed the ProCare prediction as two of the three tested and FDA-approved HIV-1 RT inhibitors indeed bind to soluble human TNF-α trimer. Interestingly, the herein disclosed similarity could be revealed neither by state-of-the-art binding sites comparison methods nor by ligand-based pairwise similarity searches, suggesting that the point cloud registration approach implemented in ProCare, is uniquely suited to identify local and unobvious similarities among totally unrelated targets.

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better.

Hundreds of fast scoring functions have been developed over the last 20 years to predict binding free energies from three-dimensional structures of protein-ligand complexes. Despite numerous statistical promises, we believe that none of them has been properly validated for daily prospective high-throughput virtual screening studies, mostly because in silico screening challenges usually employ artificially built and biased datasets. We here carry out a fully unbiased evaluation of four scoring functions (Pafnucy, ΔvinaRF20, IFP, and GRIM) on an in-house developed data collection of experimental high-confidence screening data (LIT-PCBA) covering about 3 million data points on 15 diverse pharmaceutical targets. All four scoring functions were applied to rescore the docking poses of LIT-PCBA compounds in conditions mimicking exactly standard drug discovery scenarios and were compared in terms of propensity to enrich true binders in the top 1%-ranked hit lists. Interestingly, rescoring based on simple interaction fingerprints or interaction graphs outperforms state-of-the-art machine learning and deep learning scoring functions in most of the cases. The current study notably highlights the strong tendency of deep learning methods to predict affinity values within a very narrow range centered on the mean value of samples used for training. Moreover, it suggests that knowledge of pre-existing binding modes is the key to detecting the most potent binders.

Design, Synthesis and Biological Evaluation of Arylpyridin-2-yl Guanidine Derivatives and Cyclic Mimetics as Novel MSK1 Inhibitors. An Application in an Asthma Model.

Mitogen- and Stress-Activated Kinase 1 (MSK1) is a nuclear kinase, taking part in the activation pathway of the pro-inflammatory transcription factor NF-kB and is demonstrating a therapeutic target potential in inflammatory diseases such as asthma, psoriasis and atherosclerosis. To date, few MSK1 inhibitors were reported. In order to identify new MSK1 inhibitors, a screening of a library of low molecular weight compounds was performed, and the results highlighted the 6-phenylpyridin-2-yl guanidine (compound 1a, IC50~18 µM) as a starting hit for structure-activity relationship study. Derivatives, homologues and rigid mimetics of 1a were designed, and all synthesized compounds were evaluated for their inhibitory activity towards MSK1. Among them, the non-cytotoxic 2-aminobenzimidazole 49d was the most potent at inhibiting significantly: (i) MSK1 activity, (ii) the release of IL-6 in inflammatory conditions in vitro (IC50~2 µM) and (iii) the inflammatory cell recruitment to the airways in a mouse model of asthma.

Non-Carbohydrate Glycomimetics as Inhibitors of Calcium(II)-Binding Lectins.

Because of the antimicrobial resistance crisis, lectins are considered novel drug targets. Pseudomonas aeruginosa utilizes LecA and LecB in the infection process. Inhibition of both lectins with carbohydrate-derived molecules can reduce biofilm formation to restore antimicrobial susceptibility. Here, we focused on non-carbohydrate inhibitors for LecA to explore new avenues for lectin inhibition. From a screening cascade we obtained one experimentally confirmed hit, a catechol, belonging to the well-known PAINS compounds. Rigorous analyses validated electron-deficient catechols as millimolar LecA inhibitors. The first co-crystal structure of a non-carbohydrate inhibitor in complex with a bacterial lectin clearly demonstrates the catechol mimicking the binding of natural glycosides with LecA. Importantly, catechol 3 is the first non-carbohydrate lectin ligand that binds bacterial and mammalian calcium(II)-binding lectins, giving rise to this fundamentally new class of glycomimetics.

Benchmarking Data Sets from PubChem BioAssay Data: Current Scenario and Room for Improvement.

Developing realistic data sets for evaluating virtual screening methods is a task that has been tackled by the cheminformatics community for many years. Numerous artificially constructed data collections were developed, such as DUD, DUD-E, or DEKOIS. However, they all suffer from multiple drawbacks, one of which is the absence of experimental results confirming the impotence of presumably inactive molecules, leading to possible false negatives in the ligand sets. In light of this problem, the PubChem BioAssay database, an open-access repository providing the bioactivity information of compounds that were already tested on a biological target, is now a recommended source for data set construction. Nevertheless, there exist several issues with the use of such data that need to be properly addressed. In this article, an overview of benchmarking data collections built upon experimental PubChem BioAssay input is provided, along with a thorough discussion of noteworthy issues that one must consider during the design of new ligand sets from this database. The points raised in this review are expected to guide future developments in this regard, in hopes of offering better evaluation tools for novel in silico screening procedures.

A Computer Vision Approach to Align and Compare Protein Cavities: Application to Fragment-Based Drug Design.

Identifying local similarities in binding sites from distant proteins is a major hurdle to rational drug design. We herewith present a novel method, borrowed from computer vision, adapted to mine fragment subpockets and compare them to whole ligand-binding sites. Pockets are represented by pharmacophore-annotated point clouds mimicking ideal ligands or fragments. Point cloud registration is used to find the transformation enabling an optimal overlap of points sharing similar topological and pharmacophoric neighborhoods. The method (ProCare) was calibrated on a large set of druggable cavities and applied to the comparison of fragment subpockets to entire cavities. A collection of 33,953 subpockets annotated with their bound fragments was screened for local similarity to cavities from recently described protein X-ray structures. ProCare was able to detect local similarities between remote pockets and transfer the corresponding fragments to the query cavity space, thereby proposing a first step to fragment-based design approaches targeting orphan cavities.

LIT-PCBA: An Unbiased Data Set for Machine Learning and Virtual Screening.

Comparative evaluation of virtual screening methods requires a rigorous benchmarking procedure on diverse, realistic, and unbiased data sets. Recent investigations from numerous research groups unambiguously demonstrate that artificially constructed ligand sets classically used by the community (e.g., DUD, DUD-E, MUV) are unfortunately biased by both obvious and hidden chemical biases, therefore overestimating the true accuracy of virtual screening methods. We herewith present a novel data set (LIT-PCBA) specifically designed for virtual screening and machine learning. LIT-PCBA relies on 149 dose-response PubChem bioassays that were additionally processed to remove false positives and assay artifacts and keep active and inactive compounds within similar molecular property ranges. To ascertain that the data set is suited to both ligand-based and structure-based virtual screening, target sets were restricted to single protein targets for which at least one X-ray structure is available in complex with ligands of the same phenotype (e.g., inhibitor, inverse agonist) as that of the PubChem active compounds. Preliminary virtual screening on the 21 remaining target sets with state-of-the-art orthogonal methods (2D fingerprint similarity, 3D shape similarity, molecular docking) enabled us to select 15 target sets for which at least one of the three screening methods is able to enrich the top 1%-ranked compounds in true actives by at least a factor of 2. The corresponding ligand sets (training, validation) were finally unbiased by the recently described asymmetric validation embedding (AVE) procedure to afford the LIT-PCBA data set, consisting of 15 targets and 7844 confirmed active and 407,381 confirmed inactive compounds. The data set mimics experimental screening decks in terms of hit rate (ratio of active to inactive compounds) and potency distribution. It is available online at http://drugdesign.unistra.fr/LIT-PCBA for download and for benchmarking novel virtual screening methods, notably those relying on machine learning.

Exhaustive Repertoire of Druggable Cavities at Protein-Protein Interfaces of Known Three-Dimensional Structure.

Protein-protein interactions (PPIs) offer the unique opportunity to tailor ligands aimed at specifically stabilizing or disrupting the corresponding interfaces and providing a safer alternative to conventional ligands targeting monomeric macromolecules. Selecting biologically relevant protein-protein interfaces for either stabilization or disruption by small molecules is usually biology-driven on a case-by-case basis and does not follow a structural rationale that could be applied to an entire interactome. We herewith provide a first step to the latter goal by using a fully automated and structure-based workflow, applicable to any PPI of known three-dimensional (3D) structure, to identify and prioritize druggable cavities at and nearby PPIs of pharmacological interest. When applied to the entire Protein Data Bank, 164 514 druggable cavities were identified and classified in four groups (interfacial, rim, allosteric, orthosteric) according to their properties and spatial locations. Systematic comparison of PPI cavities with pockets deduced from druggable protein-ligand complexes shows almost no overlap in property space, suggesting that even the most druggable PPI cavities are unlikely to be addressed with conventional drug-like compound libraries. The archive is freely accessible at http://drugdesign.unistra.fr/ppiome .

Unsupervised Classification of G-Protein Coupled Receptors and Their Conformational States Using IChem Intramolecular Interaction Patterns.

Over the past decade, the ever-growing structural information on G-protein coupled receptors (GPCRs) has revealed the three-dimensional (3D) characteristics of a receptor structure that is competent for G-protein binding. Structural markers are now commonly used to distinguish GPCR functional states, especially when analyzing molecular dynamics simulations. In particular, the position of the sixth helix within the seven transmembrane domains (TMs) is directly related to the coupling of the G-protein. Here, we show that the structural pattern defined by transmembrane intramolecular interactions (hydrogen bonds excluding backbone/backbone interactions, ionic bonds and aromatic interactions) is suitable for comparison of GPCR 3D structures and unsupervised distinction of the receptor states. First, we analyze a microsecond long molecular dynamic simulation of the human ß2-adrenergic receptor (ADRB2). Clustering of the 3D structures by pattern similarity identifies stable states which match the conformational classes defined by structural markers. Furthermore, the method directly spots the few state-specific interactions. Transforming pattern into graph, we extend the method to the comparison of different GPCRs. Clustering all GPCR experimentally determined structures by clique relative size first separates receptors, then their conformational states, thereby suggesting that the interaction patterns are specific of the receptor sequence and that the interaction signatures of conformational states are not shared across distant homologues.

Local Interaction Density (LID), a Fast and Efficient Tool to Prioritize Docking Poses.

Ligand docking at a protein site can be improved by prioritizing poses by similarity to validated binding modes found in the crystal structures of ligand/protein complexes. The interactions formed in the predicted model are searched in each of the reference 3D structures, taken individually. We propose to merge the information provided by all references, creating a single representation of all known binding modes. The method is called LID, an acronym for Local Interaction Density. LID was benchmarked in a pose prediction exercise on 19 proteins and 1382 ligands using PLANTS as docking software. It was also tested in a virtual screening challenge on eight proteins, with a dataset of 140,000 compounds from DUD-E and PubChem. LID significantly improved the performance of the docking program in both pose prediction and virtual screening. The gain is comparable to that obtained with a rescoring approach based on the individual comparison of reference binding modes (the GRIM method). Importantly, LID is effective with a small number of references. LID calculation time is negligible compared to the docking time.

Novel auristatin E-based albumin-binding prodrugs with superior anticancer efficacy in vivo compared to the parent compound.

Auristatins are a class of highly cytotoxic tubulin-disrupting peptides, which have shown limited therapeutic effect as free agents in clinical trials. In our continuing effort to develop acid-sensitive albumin-binding anticancer drugs exploiting circulating serum albumin as the drug carrier, we investigated the highly toxic drug payload auristatin E to assess whether the corresponding albumin-binding prodrugs were a viable option for achieving significant and concomitant tolerable antitumor activity. To achieve our goal, we developed a new aromatic maleimide-bearing linker (Sulf07) which enhanced both water solubility and stability of the prodrugs. In this study, we describe two auristatin E-based albumin-binding drugs, AE-Keto-Sulf07 and AE-Ester-Sulf07, which were designed to release the active compound at the tumor site in a pH-dependent manner. These prodrugs incorporate an acid-sensitive hydrazone bond, formed by the reaction of a carbonyl-containing auristatin E derivative with the hydrazide group of the water-solubilizing maleimide-bearing linker Sulf07. A panel of patient- and cell-derived human tumor xenograft models (melanoma A375, ovarian carcinoma A2780, non-small-cell lung cancer LXFA737 and LXFE937, and head and neck squamous cell carcinomas) were screened with starting tumor volumes in the range of either 130-150 mm3 (small tumors) or 270-380 mm3 (large tumors). Both albumin-binding prodrugs showed compelling anticancer efficacy compared to the parent drug auristatin E, inducing statistically significant long-term partial and/or complete tumor regressions. AE-Keto-Sulf07 displayed very good antitumor response over a wide dose range, 3.0-6.5 mg/kg (5-8 injections, biweekly). AE-Ester-Sulf07 was highly efficacious between 1.9 and 2.4 mg/kg (8 injections, biweekly) or at 3.8 mg/kg (4 injections, weekly), but caused cumulative skin irritation due to scratching and biting. In contrast at its MTD, auristatin E (0.3 mg/kg, 8 injections, biweekly) was only marginally active. In summary, AE-Keto-Sulf07 and AE-Ester-Sulf07 are novel acid-sensitive albumin-binding prodrugs demonstrating tumor regressions in all of the evaluated human tumor xenograft models thus supporting the stratagem that albumin can be used as an effective drug carrier for the highly potent class of auristatins.

All in One: Cavity Detection, Druggability Estimate, Cavity-Based Pharmacophore Perception, and Virtual Screening.

Discovering the very first ligands of pharmacologically important targets in a fast and cost-efficient manner is an important issue in drug discovery. In the absence of structural information on either endogenous or synthetic ligands, computational chemists classically identify the very first hits by docking compound libraries to a binding site of interest, with well-known biases arising from the usage of scoring functions. We herewith propose a novel computational method tailored to ligand-free protein structures and consisting in the generation of simple cavity-based pharmacophores to which potential ligands could be aligned by the use of a smooth Gaussian function. The method, embedded in the IChem toolkit, automatically detects ligand-binding cavities, then predicts their structural druggability, and last creates a structure-based pharmacophore for predicted druggable binding sites. A companion tool (Shaper2) was designed to align ligands to cavity-derived pharmacophoric features. The proposed method is as efficient as state-of-the-art virtual screening methods (ROCS, Surflex-Dock) in both posing and virtual screening challenges. Interestingly, IChem-Shaper2 is clearly orthogonal to these latter methods in retrieving unique chemotypes from high-throughput virtual screening data.