Synthon-based ligand discovery in virtual libraries of over 11 billion compounds.

Structure-based virtual ligand screening is emerging as a key paradigm for early drug discovery owing to the availability of high-resolution target structures1-4 and ultra-large libraries of virtual compounds5,6. However, to keep pace with the rapid growth of virtual libraries, such as readily available for synthesis (REAL) combinatorial libraries7, new approaches to compound screening are needed8,9. Here we introduce a modular synthon-based approach-V-SYNTHES-to perform hierarchical structure-based screening of a REAL Space library of more than 11 billion compounds. V-SYNTHES first identifies the best scaffold-synthon combinations as seeds suitable for further growth, and then iteratively elaborates these seeds to select complete molecules with the best docking scores. This hierarchical combinatorial approach enables the rapid detection of the best-scoring compounds in the gigascale chemical space while performing docking of only a small fraction (<0.1%) of the library compounds. Chemical synthesis and experimental testing of novel cannabinoid antagonists predicted by V-SYNTHES demonstrated a 33% hit rate, including 14 submicromolar ligands, substantially improving over a standard virtual screening of the Enamine REAL diversity subset, which required approximately 100 times more computational resources. Synthesis of selected analogues of the best hits further improved potencies and affinities (best inhibitory constant (Ki) = 0.9 nM) and CB2/CB1 selectivity (50-200-fold). V-SYNTHES was also tested on a kinase target, ROCK1, further supporting its use for lead discovery. The approach is easily scalable for the rapid growth of combinatorial libraries and potentially adaptable to any docking algorithm.

Structures of the σ2 receptor enable docking for bioactive ligand discovery.

The σ2 receptor has attracted intense interest in cancer imaging1, psychiatric disease2, neuropathic pain3-5 and other areas of biology6,7. Here we determined the crystal structure of this receptor in complex with the clinical candidate roluperidone2 and the tool compound PB288. These structures templated a large-scale docking screen of 490 million virtual molecules, of which 484 compounds were synthesized and tested. We identified 127 new chemotypes with affinities superior to 1 µM, 31 of which had affinities superior to 50 nM. The hit rate fell smoothly and monotonically with docking score. We optimized three hits for potency and selectivity, and achieved affinities that ranged from 3 to 48 nM, with up to 250-fold selectivity versus the σ1 receptor. Crystal structures of two ligands bound to the σ2 receptor confirmed the docked poses. To investigate the contribution of the σ2 receptor in pain, two potent σ2-selective ligands and one potent σ1/σ2 non-selective ligand were tested for efficacy in a mouse model of neuropathic pain. All three ligands showed time-dependent decreases in mechanical hypersensitivity in the spared nerve injury model9, suggesting that the σ2 receptor has a role in nociception. This study illustrates the opportunities for rapid discovery of in vivo probes through structure-based screens of ultra large libraries, enabling study of underexplored areas of biology.

An open-source drug discovery platform enables ultra-large virtual screens.

On average, an approved drug currently costs US$2-3 billion and takes more than 10 years to develop1. In part, this is due to expensive and time-consuming wet-laboratory experiments, poor initial hit compounds and the high attrition rates in the (pre-)clinical phases. Structure-based virtual screening has the potential to mitigate these problems. With structure-based virtual screening, the quality of the hits improves with the number of compounds screened2. However, despite the fact that large databases of compounds exist, the ability to carry out large-scale structure-based virtual screening on computer clusters in an accessible, efficient and flexible manner has remained difficult. Here we describe VirtualFlow, a highly automated and versatile open-source platform with perfect scaling behaviour that is able to prepare and efficiently screen ultra-large libraries of compounds. VirtualFlow is able to use a variety of the most powerful docking programs. Using VirtualFlow, we prepared one of the largest and freely available ready-to-dock ligand libraries, with more than 1.4 billion commercially available molecules. To demonstrate the power of VirtualFlow, we screened more than 1 billion compounds and identified a set of structurally diverse molecules that bind to KEAP1 with submicromolar affinity. One of the lead inhibitors (iKeap1) engages KEAP1 with nanomolar affinity (dissociation constant (Kd) = 114 nM) and disrupts the interaction between KEAP1 and the transcription factor NRF2. This illustrates the potential of VirtualFlow to access vast regions of the chemical space and identify molecules that bind with high affinity to target proteins.

Iterative computational design and crystallographic screening identifies potent inhibitors targeting the Nsp3 macrodomain of SARS-CoV-2.

The nonstructural protein 3 (NSP3) of the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) contains a conserved macrodomain enzyme (Mac1) that is critical for pathogenesis and lethality. While small-molecule inhibitors of Mac1 have great therapeutic potential, at the outset of the COVID-19 pandemic, there were no well-validated inhibitors for this protein nor, indeed, the macrodomain enzyme family, making this target a pharmacological orphan. Here, we report the structure-based discovery and development of several different chemical scaffolds exhibiting low- to sub-micromolar affinity for Mac1 through iterations of computer-aided design, structural characterization by ultra-high-resolution protein crystallography, and binding evaluation. Potent scaffolds were designed with in silico fragment linkage and by ultra-large library docking of over 450 million molecules. Both techniques leverage the computational exploration of tangible chemical space and are applicable to other pharmacological orphans. Overall, 160 ligands in 119 different scaffolds were discovered, and 153 Mac1-ligand complex crystal structures were determined, typically to 1 Å resolution or better. Our analyses discovered selective and cell-permeable molecules, unexpected ligand-mediated conformational changes within the active site, and key inhibitor motifs that will template future drug development against Mac1.

SNAr or Sulfonylation? Chemoselective Amination of Halo(het)arene Sulfonyl Halides for Synthetic Applications and Ultralarge Compound Library Design.

The chemoselectivity of halo(het)arene sulfonyl halide aminations is studied thoroughly under parallel synthesis conditions, and the scope and limitations of the method are established. It is shown that SNAr-reactive sulfonyl halides typically undergo sulfonamide synthesis during the first step; the second amination is also possible provided that the SNAr-active center is sufficiently reactive. On the contrary, sulfonyl fluorides bearing an arylating moiety undergo selective transformation at the latter reactive center under proper control. Further sulfur-fluoride exchange (SuFEx) is also possible, which can be especially valuable for some sulfonyl halide classes. The developed two-step parallel double amination protocol provides access to a 6.67-billion compound synthetically tractable REAL-type chemical space (76% expected synthesis success rate).

Generation of multimillion chemical space based on the parallel Groebke-Blackburn-Bienaymé reaction.

Parallel Groebke-Blackburn-Bienaymé reaction was evaluated as a source of multimillion chemically accessible chemical space. Two most popular classical protocols involving the use of Sc(OTf)3 and TsOH as the catalysts were tested on a broad substrate scope, and prevalence of the first method was clearly demonstrated. Furthermore, the scope and limitations of the procedure were established. A model 790-member library was obtained with 85% synthesis success rate. These results were used to generate a 271-Mln. readily accessible (REAL) heterocyclic chemical space mostly containing unique chemotypes, which was confirmed by comparative analysis with commercially available compound collections. Meanwhile, this chemical space contained 432 compounds that already showed biological activity according to the ChEMBL database.

Advancing molecular graphs with descriptors for the prediction of chemical reaction yields.

Chemical yield is the percentage of the reactants converted to the desired products. Chemists use predictive algorithms to select high-yielding reactions and score synthesis routes, saving time and reagents. This study suggests a novel graph neural network architecture for chemical yield prediction. The network combines structural information about participants of the transformation as well as molecular and reaction-level descriptors. It works with incomplete chemical reactions and generates reactants-product atom mapping. We show that the network benefits from advanced information by comparing it with several machine learning models and molecular representations. Models included logistic regression, support vector machine, CatBoost, and Bidirectional Encoder Representations from Transformers. Molecular representations included extended-connectivity fingerprints, Morgan fingerprints, SMILESVec embeddings, and textual. Classification and regression objectives were assessed for each model and feature set. The goal of each classification model was to separate zero- and non-zero-yielding reactions. The models were trained and evaluated on a proprietary dataset of 10 reaction types. Also, the models were benchmarked on two public single reaction type datasets. The study was supplemented with analysis of data, results, and errors, as well as the impact of steric factors, side reactions, isolation, and purification efficiency. The supplementary code is available at https://github.com/SoftServeInc/yield-paper.

A stereochemical journey around spirocyclic glutamic acid analogs.

A practical divergent synthetic approach is reported for the library of regio- and stereoisomers of glutamic acid analogs built on the spiro[3.3]heptane scaffold. Formation of the spirocyclic scaffold was achieved starting from a common precursor - an O-silylated 2-(hydroxymethyl)cyclobutanone derivative. Its olefination required using the titanium-based Tebbe protocol since the standard Wittig reaction did not work with this particular substrate. The construction of the second cyclobutane ring of the spirocyclic system was achieved through either subsequent dichloroketene addition or Meinwald oxirane rearrangement as the key synthetic steps, depending on the substitution patterns in the target compounds (1,6- or 1,5-, respectively). Further modified Strecker reaction of the resulting racemic spirocyclic ketones with the Ellman's sulfinamide as a chiral auxiliary had low to moderate diastereoselectivity; nevertheless, all stereoisomers were isolated in pure form via chromatographic separation, and their absolute configuration was confirmed by X-ray crystallography. Members of the library were tested for the inhibitory activity against H. pylori glutamate racemase.

One-pot parallel synthesis of 1,3,5-trisubstituted 1,2,4-triazoles.

An implementation of the three-component one-pot approach to unsymmetrical 1,3,5-trisubstituted-1,2,4-triazoles into combinatorial chemistry is described. The procedure is based on the coupling of amidines with carboxylic acids and subsequent cyclization with hydrazines. After the preliminary assessment of the reagent scope, the method had 81% success rate in parallel synthesis. It was shown that over a billion-sized chemical space of readily accessible ("REAL") compounds may be generated based on the proposed methodology. Analysis of physicochemical parameters shows that the library contains significant fractions of both drug-like and "beyond-rule-of-five" members. More than 10 million of accessible compounds meet the strictest lead-likeness criteria. Additionally, 195 Mln of sp3-enriched compounds can be produced. This makes the proposed approach a valuable tool in medicinal chemistry.

Multigram Synthesis of Advanced 6,6-Difluorospiro[3.3]heptane-derived Building Blocks.

A convenient methodology for constructing 6,6-difluorospiro[3.3]heptane scaffold - a conformationally restricted isostere of gem-difluorocycloalkanes - is developed. Alarge array of novel 2-mono- and 2,2-bifunctionalized difluorospiro[3.3]heptane building blocks was obtained through the convergent synthesis strategy using a common synthetic precursor - 1,1-bis(bromomethyl)-3,3-difluorocyclobutane. The target compounds and intermediates were prepared by short reaction sequences (6-10 steps) on multigram scale (up to 0.47 kg).

Last of the gem-Difluorocycloalkanes: Synthesis and Characterization of 2,2-Difluorocyclobutyl-Substituted Building Blocks.

An efficient approach to synthesis of previously unavailable 2-substituted difluorocyclobutane building blocks was developed and applied on a multigram scale. The key step of the synthetic sequence included deoxofluorination of O-protected 2-(hydroxylmethyl)cyclobutanone. Dissociation constants (p Ka) and log P values for 2,2-difluorocyclobutaneamine and 2,2-difluorocyclobutanecarboxylic acid or their derivatives were measured and compared with the values obtained for the corresponding 3,3-difluorocyclobutane derivatives and nonfluorinated counterparts. Three-dimensional structures of 2,2- and 3,3-difluorocyclobutanamines were compared using exit vector plot analysis of X-ray crystallographic data.

Does a methionine-to-norleucine substitution in PGLa influence peptide-membrane interactions?

Yes. To understand the molecular mechanisms of amphiphilic membrane-active peptides, it is essential to study their interactions with lipid bilayers under near-native conditions. Amino acid composition largely determines the non-specific properties of peptides, on the basis of the physicochemical properties of the side chains. The resultant effects on peptides' functional properties include influences on the conformation, structural dynamics and binding affinities within the peptide interactome. Here, we studied the effect of substituting oxidation-prone methionine (Met) with non-oxidizable norleucine (Nle) in the model α-helical antimicrobial peptide PGLa, through systematic comparison of PGLa with the (2)Met/(2)Nle mutant. Both peptides were evaluated for their bacteriostatic and hemolytic activities (using in situ assays), for their conformational preferences in isotropic solutions (using circular dichroism spectropolarimetry) and for their abilities to modulate membrane curvature (using a solid-state (31)P NMR assay). We determined the membrane-bound states in detail and characterized the orientational dynamics of both peptides in oriented phospholipid membranes by solid-state (19)F NMR spectroscopy. On the one hand, the bioactivity results, the structure in the diluted membrane-mimicking environments and the strong inhibition of the negative membrane curvature were comparable between PGLa and the mutant. On the other hand, the alignments in DMPC bilayer were qualitatively the same but differed in absolute values - the more hydrophobic Nle residue inserted deeper in the membrane core. Furthermore, the mutant peptide displayed a significantly reduced ability to re-orient from the monomeric, surficial to the putative dimeric, tilted state. Overall, these results confirm the functional isosterism of Nle and Met in the helical membrane-active peptides but highlight differences in the ways in which the two residues affect non-specific binding to the lipid bilayer and homomeric peptide-peptide interactions.

Delivering Structural Information on the Polar Face of Membrane-Active Peptides: 19 F-NMR Labels with a Cationic Side Chain.

Conformationally constrained non-racemizing trifluoromethyl-substituted lysine isosteres [(E)- and (Z)-TCBLys] with charged side chains are presented as a new type of 19 F-NMR labels for peptide studies. Design of the labels, their synthesis, incorporation into peptides and experimental demonstration of their application for solid state NMR studies of membrane-active peptides are described. A series of fluorine-labeled analogues of the helical amphipathic antimicrobial peptide PGLa(Nle) was obtained, in which different lysine residues in the original peptide sequence were replaced, one at a time, by either (E)- or (Z)-TCBLys. Antimicrobial activities of the synthesized analogues were practically the same as those of the parent peptide. The structural and orientational parameters of the helical PGLa(Nle) peptide in model bilayers, as determined using the novel labels confirmed and refined the previously known structure. (E)- and (Z)-TCBLys, as a set of cationic 19 F-NMR labels, were shown to deliver structural information about the charged face of amphipathic peptides by solid state 19 F-NMR, previously inaccessible by this method.

Synthesis and structural analysis of angular monoprotected diamines based on spiro[3.3]heptane scaffold.

The synthesis of all stereoisomers of spiro[3.3]heptane-1,6-diamines suitably protected for use as building blocks in drug discovery is reported. Structural analysis revealed the similarity between the spiro[3.3]heptane and cyclohexane scaffolds. Comparison of the distance between functional groups and their spatial orientation proved that (1S,4r,6R)- and (1R,4r,6S)-1,6-disubstituted spiro[3.3]heptanes can be considered as restricted surrogates of cis-1,4-disubstituted cyclohexane derivatives. Similarly, (1S,4s,6R)- and (1R,4s,6S)-1,6-disubstituted spiro[3.3]heptanes are the restricted surrogates of trans-1,3-disubstituted cyclohexanes. Such replacement can be recommended for use in optimization of ADME parameters of lead compounds in drug discovery.

The impact of Library Size and Scale of Testing on Virtual Screening.

Virtual libraries for ligand discovery have recently increased 10,000-fold, and this is thought to have improved hit rates and potencies from library docking. This idea has not, however, been experimentally tested in direct comparisons of larger-vs-smaller libraries. Meanwhile, though libraries have exploded, the scale of experimental testing has little changed, with often only dozens of high-ranked molecules investigated, making interpretation of hit rates and affinities uncertain. Accordingly, we docked a 1.7 billion molecule virtual library against the model enzyme AmpC ß-lactamase, testing 1,521 new molecules and comparing the results to the same screen with a library of 99 million molecules, where only 44 molecules were tested. Encouragingly, the larger screen outperformed the smaller one: hit rates improved by two-fold, more new scaffolds were discovered, and potency improved. Overall, 50-fold more inhibitors were found, supporting the idea that there are many more compounds to be discovered than are being tested. With so many compounds evaluated, we could ask how the results vary with number tested, sampling smaller sets at random from the 1521. Hit rates and affinities were highly variable when we only sampled dozens of molecules, and it was only when we included several hundred molecules that results converged. As docking scores improved, so too did the likelihood of a molecule binding; hit rates improved steadily with docking score, as did affinities. This also appeared true on reanalysis of large-scale results against the σ2 and dopamine D4 receptors. It may be that as the scale of both the virtual libraries and their testing grows, not only are better ligands found but so too does our ability to rank them.

Large library docking for cannabinoid-1 receptor agonists with reduced side effects.

Large library docking can reveal unexpected chemotypes that complement the structures of biological targets. Seeking new agonists for the cannabinoid-1 receptor (CB1R), we docked 74 million tangible molecules, prioritizing 46 high ranking ones for de novo synthesis and testing. Nine were active by radioligand competition, a 20% hit-rate. Structure-based optimization of one of the most potent of these (Ki = 0.7 uM) led to '4042, a 1.9 nM ligand and a full CB1R agonist. A cryo-EM structure of the purified enantiomer of '4042 ('1350) in complex with CB1R-Gi1 confirmed its docked pose. The new agonist was strongly analgesic, with generally a 5-10-fold therapeutic window over sedation and catalepsy and no observable conditioned place preference. These findings suggest that new cannabinoid chemotypes may disentangle characteristic cannabinoid side-effects from their analgesia, supporting the further development of cannabinoids as pain therapeutics.

Generative and reinforcement learning approaches for the automated de novo design of bioactive compounds.

Deep generative neural networks have been used increasingly in computational chemistry for de novo design of molecules with desired properties. Many deep learning approaches employ reinforcement learning for optimizing the target properties of the generated molecules. However, the success of this approach is often hampered by the problem of sparse rewards as the majority of the generated molecules are expectedly predicted as inactives. We propose several technical innovations to address this problem and improve the balance between exploration and exploitation modes in reinforcement learning. In a proof-of-concept study, we demonstrate the application of the deep generative recurrent neural network architecture enhanced by several proposed technical tricks to design inhibitors of the epidermal growth factor (EGFR) and further experimentally validate their potency. The proposed technical solutions are expected to substantially improve the success rate of finding novel bioactive compounds for specific biological targets using generative and reinforcement learning approaches.

Synthesis of 2-azaspiro[3.3]heptane-derived amino acids: ornitine and GABA analogues.

Synthesis of 6-amino-2-azaspiro[3.3]heptane-6-carboxylic acid and 2-azaspiro[3.3]heptane-6-carboxylic acid was performed. Both four-membered rings in the spirocyclic scaffold were constructed by subsequent ring closure of corresponding 1,3-bis-electrophiles at 1,1-C- or 1,1-N-bis-nucleophiles. The two novel amino acids were added to the family of the sterically constrained amino acids for the use in chemistry, biochemistry, and drug design.

Cyclobutane-derived diamines: synthesis and molecular structure.

Cyclobutane diamines (i.e., cis- and trans-1,3-diaminocyclobutane, 6-amino-3-azaspiro[3.3]heptane, and 3,6-diaminospiro[3.3]heptane) are considered as promising sterically constrained diamine building blocks for drug discovery. An approach to the syntheses of their Boc-monoprotected derivatives has been developed aimed at the preparation of multigram amounts of the compounds. These novel synthetic schemes exploit classical malonate alkylation chemistry for the construction of cyclobutane rings. The conformational preferences of the cyclobutane diamine derivatives have been evaluated by X-ray diffraction and compared with the literature data on sterically constrained diamines, which are among the constituents of commercially available drugs.

Erratum: Generating Multibillion Chemical Space of Readily Accessible Screening Compounds.

[This corrects the article DOI: 10.1016/j.isci.2020.101681.].