SCUBIDOO: A Large yet Screenable and Easily Searchable Database of Computationally Created Chemical Compounds Optimized toward High Likelihood of Synthetic Tractability

A virtual library of small molecules mimicking dipeptides

Virtual combinatorial libraries are prevalent in drug discovery due to improvements in the prediction of synthetic reactions that can be performed. This has gone hand in hand with the development of virtual screening capabilities to effectively screen the large chemical spaces spanned by exhaustive enumeration of reaction products. In this study, we generated a small-molecule dipeptide mimic library to target proteins binding small peptides. The library was created based on the general idea of peptide synthesis, that is, amino acid mimics were reacted in silico to form the dipeptide mimics, yielding 2,036,819 unique compounds. After docking calculations, two compounds from the library were synthesized and tested against WD repeat-containing protein 5 (WDR5) and histamine receptors H1-H4 to evaluate whether these molecules are viable in assays. The compounds showed the highest potency at the histamine H3 receptor, with Ki values in the two-digit micromolar range.

ChemoDOTS: a web server to design chemistry-driven focused libraries

In drug discovery, the successful optimization of an initial hit compound into a lead molecule requires multiple cycles of chemical modification. Consequently, there is a need to efficiently generate synthesizable chemical libraries to navigate the chemical space surrounding the primary hit. To address this need, we introduce ChemoDOTS, an easy-to-use web server for hit-to-lead chemical optimization freely available at https://chemodots.marseille.inserm.fr/. With this tool, users enter an activated form of the initial hit molecule then choose from automatically detected reactive functions. The server proposes compatible chemical transformations via an ensemble of encoded chemical reactions widely used in the pharmaceutical industry during hit-to-lead optimization. After selection of the desired reactions, all compatible chemical building blocks are automatically coupled to the initial hit to generate a raw chemical library. Post-processing filters can be applied to extract a subset of compounds with specific physicochemical properties. Finally, explicit stereoisomers and tautomers are computed, and a 3D conformer is generated for each molecule. The resulting virtual library is compatible with most docking software for virtual screening campaigns. ChemoDOTS rapidly generates synthetically feasible, hit-focused, large, diverse chemical libraries with finely-tuned physicochemical properties via a user-friendly interface providing a powerful resource for researchers engaged in hit-to-lead optimization.

DFRscore: Deep Learning-Based Scoring of Synthetic Complexity with Drug-Focused Retrosynthetic Analysis for High-Throughput Virtual Screening

Recently emerging generative AI models enable us to produce a vast number of compounds for potential applications. While they can provide novel molecular structures, the synthetic feasibility of the generated molecules is often questioned. To address this issue, a few recent studies have attempted to use deep learning models to estimate the synthetic accessibility of many molecules rapidly. However, retrosynthetic analysis tools used to train the models rely on reaction templates automatically extracted from a large reaction database that are not domain-specific and may exhibit low chemical correctness. To overcome this limitation, we introduce DFRscore (Drug-Focused Retrosynthetic score), a deep learning-based approach for a more practical assessment of synthetic accessibility in drug discovery. The DFRscore model is trained exclusively on drug-focused reactions, providing a predicted number of minimally required synthetic steps for each compound. This approach enables practitioners to filter out compounds that do not meet their desired level of synthetic accessibility at an early stage of high-throughput virtual screening for accelerated drug discovery. The proposed strategy can be easily adapted to other domains by adjusting the synthesis planning setup of the reaction templates and starting materials.

SpaceGrow: efficient shape-based virtual screening of billion-sized combinatorial fragment spaces

The growing size of make-on-demand chemical libraries is posing new challenges to cheminformatics. These ultra-large chemical libraries became too large for exhaustive enumeration. Using a combinatorial approach instead, the resource requirement scales approximately with the number of synthons instead of the number of molecules. This gives access to billions or trillions of compounds as so-called chemical spaces with moderate hardware and in a reasonable time frame. While extremely performant ligand-based 2D methods exist in this context, 3D methods still largely rely on exhaustive enumeration and therefore fail to apply. Here, we present SpaceGrow: a novel shape-based 3D approach for ligand-based virtual screening of billions of compounds within hours on a single CPU. Compared to a conventional superposition tool, SpaceGrow shows comparable pose reproduction capacity based on RMSD and superior ranking performance while being orders of magnitude faster. Result assessment of two differently sized subsets of the eXplore space reveals a higher probability of finding superior results in larger spaces highlighting the potential of searching in ultra-large spaces. Furthermore, the application of SpaceGrow in a drug discovery workflow was investigated in four examples involving G protein-coupled receptors (GPCRs) with the aim to identify compounds with similar binding capabilities and molecular novelty.

Discovery of Potential Neonicotinoid Insecticides by an Artificial Intelligence Generative Model and Structure-Based Virtual Screening

The identification of neonicotinoid insecticides bearing novel scaffolds is of great importance for pesticide discovery. Here, artificial intelligence-based tools and virtual screening strategy were integrated to discover potential leads of neonicotinoid insecticides. A deep generative model was successfully constructed using a recurrent neural network combined with transfer learning. The model evaluation showed that the pretrained model could accurately grasp the SMILES grammar of drug-like molecules and generate potential neonicotinoid compounds after transfer learning. The generated molecules were evaluated by hierarchical virtual screening, hits were subjected to a similarity search, and the most similar structures were purchased for the bioassay. Compounds A2 and A5 displayed 52.5 and 50.3% mortality rates against Aphis craccivora at 100 mg/L, respectively. The docking study indicated that these two compounds have similar binding modes to neonicotinoids, which were verified by further molecular dynamics simulations.

Constructing Innovative Covalent and Noncovalent Compound Libraries: Insights from 3D Protein-Ligand Interactions

Noncovalent interactions between small-molecule drugs and protein targets assume a pivotal role in drug design. Moreover, the design of covalent inhibitors, forming covalent bonds with amino acid residues, requires rational reactivity for their covalent warheads, presenting a key challenge as well. Understanding the intricacies of these interactions provides a more comprehensive understanding of molecular binding mechanisms, thereby guiding the rational design of potent inhibitors. In this study, we adopted the fragment-based drug design approach, introducing a novel methodology to extract noncovalent and covalent fragments according to distinct three-dimensional (3D) interaction modes from noncovalent and covalent compound libraries. Additionally, we systematically replaced existing ligands with rational fragment substitutions, based on the spatial orientation of fragments in 3D space. Furthermore, we adopted a molecular generation approach to create innovative covalent inhibitors. This process resulted in the recombination of a noncovalent compound library and several covalent compound libraries, constructed by two commonly encountered covalent amino acids: cysteine and serine. We utilized noncovalent ligands in KLIFS and covalent ligands in CovBinderInPDB as examples to recombine noncovalent and covalent libraries. These recombined compound libraries cover a substantial portion of the chemical space present in the original compound libraries and exhibit superior performance in terms of molecular scaffold diversity compared to the original compound libraries and other 11 commercial libraries. We also recombined BTK-focused libraries, and 23 compounds within our libraries have been validated by former researchers to possess potential biological activity. The establishment of these compound libraries provides valuable resources for virtual screening of covalent and noncovalent drugs targeting similar molecular targets.

Sequence-based drug design as a concept in computational drug design

Drug development based on target proteins has been a successful approach in recent decades. However, the conventional structure-based drug design (SBDD) pipeline is a complex, human-engineered process with multiple independently optimized steps. Here, we propose a sequence-to-drug concept for computational drug design based on protein sequence information by end-to-end differentiable learning. We validate this concept in three stages. First, we design TransformerCPI2.0 as a core tool for the concept, which demonstrates generalization ability across proteins and compounds. Second, we interpret the binding knowledge that TransformerCPI2.0 learned. Finally, we use TransformerCPI2.0 to discover new hits for challenging drug targets, and identify new target for an existing drug based on an inverse application of the concept. Overall, this proof-of-concept study shows that the sequence-to-drug concept adds a perspective on drug design. It can serve as an alternative method to SBDD, particularly for proteins that do not yet have high-quality 3D structures available.

FastGrow: on-the-fly growing and its application to DYRK1A

Fragment-based drug design is an established routine approach in both experimental and computational spheres. Growing fragment hits into viable ligands has increasingly shifted into the spotlight. FastGrow is an application based on a shape search algorithm that addresses this challenge at high speeds of a few milliseconds per fragment. It further features a pharmacophoric interaction description, ensemble flexibility, as well as geometry optimization to become a fully fledged structure-based modeling tool. All features were evaluated in detail on a previously reported collection of fragment growing scenarios extracted from crystallographic data. FastGrow was also shown to perform competitively versus established docking software. A case study on the DYRK1A kinase, using recently reported new chemotypes, illustrates FastGrow's features in practice and its ability to identify active fragments. FastGrow is freely available to the public as a web server at https://fastgrow.plus/ and is part of the SeeSAR 3D software package.

A Transfer Learning Approach for Reaction Discovery in Small Data Situations Using Generative Model

Sustainable practices in chemical sciences can be better realized by adopting interdisciplinary approaches that combine the advantages of machine learning (ML) on the initially acquired small data in reaction discovery. Developing new reactions generally remains heuristic and even time and resource intensive. For instance, synthesis of fluorine-containing compounds, which constitute ∼20% of the marketed drugs, relies on deoxyfluorination of abundantly available alcohols. Herein, we demonstrate the use of a recurrent neural network-based deep generative model built on a library of just 37 alcohols for effective learning and exploration of the chemical space. The proof-of-concept ML model is able to generate good quality, synthetically accessible, higher-yielding novel alcohol molecules. This protocol would have superior utility for deployment into a practical reaction discovery pipeline.

•

Dual pronged transfer learning, both to generate and predict yields of new molecules

•

Demonstrated the utility for an important family of deoxyfluorination of alcohols

•

Applicable for practically more likely situations with relatively smaller data

•

Extendable to other reaction manifolds to facilitate expedited reaction discovery

Prediction of Compound Synthesis Accessibility Based on Reaction Knowledge Graph

With the increasing application of deep-learning-based generative models for de novo molecule design, the quantitative estimation of molecular synthetic accessibility (SA) has become a crucial factor for prioritizing the structures generated from generative models. It is also useful for helping in the prioritization of hit/lead compounds and guiding retrosynthesis analysis. In this study, based on the USPTO and Pistachio reaction datasets, a chemical reaction network was constructed for the identification of the shortest reaction paths (SRP) needed to synthesize compounds, and different SRP cut-offs were then used as the threshold to distinguish a organic compound as either an easy-to-synthesize (ES) or hard-to-synthesize (HS) class. Two synthesis accessibility models (DNN-ECFP model and graph-based CMPNN model) were built using deep learning/machine learning algorithms. Compared to other existing synthesis accessibility scoring schemes, such as SYBA, SCScore, and SAScore, our results show that CMPNN (ROC AUC: 0.791) performs better than SYBA (ROC AUC: 0.76), albeit marginally, and outperforms SAScore and SCScore. Our prediction models based on historical reaction knowledge could be a potential tool for estimating molecule SA.

LEADD: Lamarckian evolutionary algorithm for de novo drug design

Given an objective function that predicts key properties of a molecule, goal-directed de novo molecular design is a useful tool to identify molecules that maximize or minimize said objective function. Nonetheless, a common drawback of these methods is that they tend to design synthetically unfeasible molecules. In this paper we describe a Lamarckian evolutionary algorithm for de novo drug design (LEADD). LEADD attempts to strike a balance between optimization power, synthetic accessibility of designed molecules and computational efficiency. To increase the likelihood of designing synthetically accessible molecules, LEADD represents molecules as graphs of molecular fragments, and limits the bonds that can be formed between them through knowledge-based pairwise atom type compatibility rules. A reference library of drug-like molecules is used to extract fragments, fragment preferences and compatibility rules. A novel set of genetic operators that enforce these rules in a computationally efficient manner is presented. To sample chemical space more efficiently we also explore a Lamarckian evolutionary mechanism that adapts the reproductive behavior of molecules. LEADD has been compared to both standard virtual screening and a comparable evolutionary algorithm using a standardized benchmark suite and was shown to be able to identify fitter molecules more efficiently. Moreover, the designed molecules are predicted to be easier to synthesize than those designed by other evolutionary algorithms.

Identification of lead inhibitors of TMPRSS2 isoform 1 of SARS-CoV-2 target using neural network, random forest, and molecular docking

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was responsible for over 4 million confirmed cases of severe acute respiratory syndrome, of which more than 300,000 cases were confirmed to be dead as of May 2020. The virulent endocytotic activities of SARS-CoV-2 have been associated with angiotensin-converting enzyme 2 (ACE2) and transmembrane protease serine 2 (TMPRSS2). Previous studies on the viral activation of TMPRSS2 focused most often than not on the isoform 2 of TMPRSS2, but the isoform 1 (529 residues) has also been shown to be expressed in target cells and contribute to viral activation in host. The inhibition of TMPRSS2 has been reported to grossly reduce the pathogenic effects of SARS-CoV-2 endocytotic activities.

In this study therefore, we developed two machine learning models using random forest classifier (RFC) and neural networks (NNs) based on 2251 serine protease inhibitors to screen a database of 21,000,000 virtual compounds. We screened the hit compounds using absorption, distribution, metabolism, and excretion (ADME) properties and finally docked the filtered compounds into the predicted binding site of TMPRSS2 isoform 1 homology model to determine their corresponding binding affinity and plausible molecular interactions.

One (ASONN) and four (ASOIRFC1–4) lead compounds were obtained from the ADME-NN and RFC filtered hits, respectively, having better binding affinity and lead-likeness properties than those of camostat; this could be due to extensive hydrogen and hydrophobic interactions.

What is in Our Kit? An Analysis of Building Blocks Used in Medicinal Chemistry Parallel Libraries

Building blocks are the molecular foundations for drug molecule design. The building block is one of the determining factors of final compound qualities in any given medicinal chemistry campaign. Herein, we describe our analysis of the building blocks used in parallel library synthesis at AbbVie. The results gave insights into the synthetic tractability and accessibilities of building blocks used in medicinal chemistry. Furthermore, our analysis showed that opportunities still exist for the identification and future incorporation of underrepresented building blocks, even for commonly used reactions, to obtain intellectual and competitive advantages in drug discovery.

Predicting novel drug candidates against Covid-19 using generative deep neural networks

The novel Coronavirus outbreak has created a massive economic crisis, and many succumb to death, disturbing the lives of mankind all over the world. Currently, there are no viable treatment for this condition, drug development approaches are being pursued with vigor. The major treatment options are to repurpose existing drugs or to find new ones. Traditional methods for drug discovery take a longer time, so there is an urgent need to develop some alternative techniques that reduces search space for drug candidates. Towards this endeavor, we propose a novel drug discovery method that leverages on long short term memory (LSTM) model to generate novel molecules that are adept at binding with novel Coronavirus protease. Our study demonstrates that the proposed method is able to recreate novel molecules that correlate very much with the properties of trained molecules. Further, we fine-tune the model to generate novel drug-like molecules that are active towards a specific target. We consider 3CLPro, the main protease of novel Coronavirus, as a therapeutic target and demonstrated in silico screening to assess target structural binding affinities with docking simulations. We observed that 80% of generated molecules show docking free energy of less than −5.8 kcal/mol. The top generated drug candidate has the highest binding affinity with a docking score of −8.5 kcal/mol, which is very much lower when compared to approved existing commercial drugs including, Remdesivir. The low binding energy indicates that the generated molecules could be explored as potential drug candidates for Covid-19.

Fragment evolution for GPCRs: the role of secondary binding sites in optimization

We developed a docking-based fragment evolution approach that extends orthosteric fragments towards a less conserved secondary binding pocket of GPCRs. Evaluating 13 000 extensions for the β₁- and β₂-adrenergic receptors we synthesized and tested 112 bitopic molecules. Our results confirmed the positive contribution of the secondary binding pocket to both potency and selectivity optimizations.

Frag4Lead: growing crystallographic fragment hits by catalog using fragment-guided template docking

In recent years, crystallographic fragment screening has matured into an almost routine experiment at several modern synchrotron sites. The hits of the screening experiment, i.e. small molecules or fragments binding to the target protein, are revealed along with their 3D structural information. Therefore, they can serve as useful starting points for further structure-based hit-to-lead development. However, the progression of fragment hits to tool compounds or even leads is often hampered by a lack of chemical feasibility. As an attractive alternative, compound analogs that embed the fragment hit structurally may be obtained from commercial catalogs. Here, a workflow is reported based on filtering and assessing such potential follow-up compounds by template docking. This means that the crystallographic binding pose was integrated into the docking calculations as a central starting parameter. Subsequently, the candidates are scored on their interactions within the binding pocket. In an initial proof-of-concept study using five starting fragments known to bind to the aspartic protease endothiapepsin, 28 follow-up compounds were selected using the designed workflow and their binding was assessed by crystallography. Ten of these compounds bound to the active site and five of them showed significantly increased affinity in isothermal titration calorimetry of up to single-digit micromolar affinity. Taken together, this strategy is capable of efficiently evolving the initial fragment hits without major synthesis efforts and with full control by X-ray crystallography.

Retrosynthetic accessibility score (RAscore) - rapid machine learned synthesizability classification from AI driven retrosynthetic planning

Computer aided synthesis planning (CASP) is part of a suite of artificial intelligence (AI) based tools that are able to propose synthesis routes to a wide range of compounds. However, at present they are too slow to be used to screen the synthetic feasibility of millions of generated or enumerated compounds before identification of potential bioactivity by virtual screening (VS) workflows. Herein we report a machine learning (ML) based method capable of classifying whether a synthetic route can be identified for a particular compound or not by the CASP tool AiZynthFinder. The resulting ML models return a retrosynthetic accessibility score (RAscore) of any molecule of interest, and computes at least 4500 times faster than retrosynthetic analysis performed by the underlying CASP tool. The RAscore should be useful for pre-screening millions of virtual molecules from enumerated databases or generative models for synthetic accessibility and produce higher quality databases for virtual screening of biological activity.

Retrosynthetic accessibility score (RAscore) - rapid machine learned synthesizability classification from AI driven retrosynthetic planning

Computer aided synthesis planning (CASP) is part of a suite of artificial intelligence (AI) based tools that are able to propose synthesis routes to a wide range of compounds. However, at present they are too slow to be used to screen the synthetic feasibility of millions of generated or enumerated compounds before identification of potential bioactivity by virtual screening (VS) workflows. Herein we report a machine learning (ML) based method capable of classifying whether a synthetic route can be identified for a particular compound or not by the CASP tool AiZynthFinder. The resulting ML models return a retrosynthetic accessibility score (RAscore) of any molecule of interest, and computes at least 4500 times faster than retrosynthetic analysis performed by the underlying CASP tool. The RAscore should be useful for pre-screening millions of virtual molecules from enumerated databases or generative models for synthetic accessibility and produce higher quality databases for virtual screening of biological activity.

The retrosynthetic accessibility score (RAscore) is based on AI driven retrosynthetic planning, and is useful for rapid scoring of synthetic feasability and pre-screening of large datasets of virtual/generated molecules.

Profiling and analysis of chemical compounds using pointwise mutual information

Pointwise mutual information (PMI) is a measure of association used in information theory. In this paper, PMI is used to characterize several publicly available databases (DrugBank, ChEMBL, PubChem and ZINC) in terms of association strength between compound structural features resulting in database PMI interrelation profiles. As structural features, substructure fragments obtained by coding individual compounds as MACCS, PubChemKey and ECFP fingerprints are used. The analysis of publicly available databases reveals, in accord with other studies, unusual properties of DrugBank compounds which further confirms the validity of PMI profiling approach. Z-standardized relative feature tightness (ZRFT), a PMI-derived measure that quantifies how well the given compound's feature combinations fit these in a particular compound set, is applied for the analysis of compound synthetic accessibility (SA), as well as for the classification of compounds as easy (ES) and hard (HS) to synthesize. ZRFT value distributions are compared with these of SYBA and SAScore. The analysis of ZRFT values of structurally complex compounds in the SAVI database reveals oligopeptide structures that are mispredicted by SAScore as HS, while correctly predicted by ZRFT and SYBA as ES. Compared to SAScore, SYBA and random forest, ZRFT predictions are less accurate, though by a narrow margin (AccZRFT = 94.5%, AccSYBA = 98.8%, AccSAScore = 99.0%, AccRF = 97.3%). However, ZRFT ability to distinguish between ES and HS compounds is surprisingly high considering that while SYBA, SAScore and random forest are dedicated SA models, ZRFT is a generic measurement that merely quantifies the strength of interrelations between structural feature pairs. The results presented in the current work indicate that structural feature co-occurrence, quantified by PMI or ZRFT, contains a significant amount of information relevant to physico-chemical properties of organic compounds.

DrugSpaceX: a large screenable and synthetically tractable database extending drug space

One of the most prominent topics in drug discovery is efficient exploration of the vast drug-like chemical space to find synthesizable and novel chemical structures with desired biological properties. To address this challenge, we created the DrugSpaceX (https://drugspacex.simm.ac.cn/) database based on expert-defined transformations of approved drug molecules. The current version of DrugSpaceX contains >100 million transformed chemical products for virtual screening, with outstanding characteristics in terms of structural novelty, diversity and large three-dimensional chemical space coverage. To illustrate its practical application in drug discovery, we used a case study of discoidin domain receptor 1 (DDR1), a kinase target implicated in fibrosis and other diseases, to show DrugSpaceX performing a quick search of initial hit compounds. Additionally, for ligand identification and optimization purposes, DrugSpaceX also provides several subsets for download, including a 10% diversity subset, an extended drug-like subset, a drug-like subset, a lead-like subset, and a fragment-like subset. In addition to chemical properties and transformation instructions, DrugSpaceX can locate the position of transformation, which will enable medicinal chemists to easily integrate strategy planning and protection design.

Artificial intelligence and automation in computer aided synthesis planning

In this perspective we deal with questions pertaining to the development of synthesis planning technologies over the course of recent years.

Shape-Based Descriptors for Efficient Structure-Based Fragment Growing

Structure-based fragment growing is one of the key techniques in fragment-based drug design. Fragment growing is commonly practiced based on structural and biophysical data. Computational workflows are employed to predict which fragment elaborations could lead to high-affinity binders. Several such workflows exist but many are designed to be long running noninteractive systems. Shape-based descriptors have been proven to be fast and perform well at virtual-screening tasks. They could, therefore, be applied to the fragment-growing problem to enable an interactive fragment-growing workflow. In this work, we describe and analyze the use of specific shape-based directional descriptors for the task of fragment growing. The performance of these descriptors that we call ray volume matrices (RVMs) is evaluated on two data sets containing protein-ligand complexes. While the first set focuses on self-growing, the second measures practical performance in a cross-growing scenario. The runtime of screenings using RVMs as well as their robustness to three dimensional perturbations is also investigated. Overall, it can be shown that RVMs are useful to prefilter fragment candidates. For up to 84% of the 3299 generated self-growing cases and for up to 66% of the 326 generated cross-growing cases, RVMs could create poses with less than 2 Å root-mean-square deviation to the crystal structure with average query speeds of around 30,000 conformations per second. This opens the door for fast explorative screenings of fragment libraries.

SAVI, in silico generation of billions of easily synthesizable compounds through expert-system type rules

We have made available a database of over 1 billion compounds predicted to be easily synthesizable, called Synthetically Accessible Virtual Inventory (SAVI). They have been created by a set of transforms based on an adaptation and extension of the CHMTRN/PATRAN programming languages describing chemical synthesis expert knowledge, which originally stem from the LHASA project. The chemoinformatics toolkit CACTVS was used to apply a total of 53 transforms to about 150,000 readily available building blocks (enamine.net). Only single-step, two-reactant syntheses were calculated for this database even though the technology can execute multi-step reactions. The possibility to incorporate scoring systems in CHMTRN allowed us to subdivide the database of 1.75 billion compounds in sets according to their predicted synthesizability, with the most-synthesizable class comprising 1.09 billion synthetic products. Properties calculated for all SAVI products show that the database should be well-suited for drug discovery. It is being made publicly available for free download from https://doi.org/10.35115/37n9-5738.

Advancing computer-aided drug discovery (CADD) by big data and data-driven machine learning modeling

Advancing a new drug to market requires substantial investments in time as well as financial resources. Crucial bioactivities for drug candidates, including their efficacy, pharmacokinetics (PK), and adverse effects, need to be investigated during drug development. With advancements in chemical synthesis and biological screening technologies over the past decade, a large amount of biological data points for millions of small molecules have been generated and are stored in various databases. These accumulated data, combined with new machine learning (ML) approaches, such as deep learning, have shown great potential to provide insights into relevant chemical structures to predict in vitro, in vivo, and clinical outcomes, thereby advancing drug discovery and development in the big data era.

Phenotypic Screening of Chemical Libraries Enriched by Molecular Docking to Multiple Targets Selected from Glioblastoma Genomic Data

Like most solid tumors, glioblastoma multiforme (GBM) harbors multiple overexpressed and mutated genes that affect several signaling pathways. Suppressing tumor growth of solid tumors like GBM without toxicity may be achieved by small molecules that selectively modulate a collection of targets across different signaling pathways, also known as selective polypharmacology. Phenotypic screening can be an effective method to uncover such compounds, but the lack of approaches to create focused libraries tailored to tumor targets has limited its impact. Here, we create rational libraries for phenotypic screening by structure-based molecular docking chemical libraries to GBM-specific targets identified using the tumor's RNA sequence and mutation data along with cellular protein-protein interaction data. Screening this enriched library of 47 candidates led to several active compounds, including 1 (IPR-2025), which (i) inhibited cell viability of low-passage patient-derived GBM spheroids with single-digit micromolar IC50 values that are substantially better than standard-of-care temozolomide, (ii) blocked tube-formation of endothelial cells in Matrigel with submicromolar IC50 values, and (iii) had no effect on primary hematopoietic CD34+ progenitor spheroids or astrocyte cell viability. RNA sequencing provided the potential mechanism of action for 1, and mass spectrometry-based thermal proteome profiling confirmed that the compound engages multiple targets. The ability of 1 to inhibit GBM phenotypes without affecting normal cell viability suggests that our screening approach may hold promise for generating lead compounds with selective polypharmacology for the development of treatments of incurable diseases like GBM.

Autonomous Discovery in the Chemical Sciences Part II: Outlook

This two-part Review examines how automation has contributed to different aspects of discovery in the chemical sciences. In this second part, we reflect on a selection of exemplary studies. It is increasingly important to articulate what the role of automation and computation has been in the scientific process and how that has or has not accelerated discovery. One can argue that even the best automated systems have yet to "discover" despite being incredibly useful as laboratory assistants. We must carefully consider how they have been and can be applied to future problems of chemical discovery in order to effectively design and interact with future autonomous platforms. The majority of this Review defines a large set of open research directions, including improving our ability to work with complex data, build empirical models, automate both physical and computational experiments for validation, select experiments, and evaluate whether we are making progress towards the ultimate goal of autonomous discovery. Addressing these practical and methodological challenges will greatly advance the extent to which autonomous systems can make meaningful discoveries.

SYBA: Bayesian estimation of synthetic accessibility of organic compounds

SYBA (SYnthetic Bayesian Accessibility) is a fragment-based method for the rapid classification of organic compounds as easy- (ES) or hard-to-synthesize (HS). It is based on a Bernoulli naïve Bayes classifier that is used to assign SYBA score contributions to individual fragments based on their frequencies in the database of ES and HS molecules. SYBA was trained on ES molecules available in the ZINC15 database and on HS molecules generated by the Nonpher methodology. SYBA was compared with a random forest, that was utilized as a baseline method, as well as with other two methods for synthetic accessibility assessment: SAScore and SCScore. When used with their suggested thresholds, SYBA improves over random forest classification, albeit marginally, and outperforms SAScore and SCScore. However, upon the optimization of SAScore threshold (that changes from 6.0 to – 4.5), SAScore yields similar results as SYBA. Because SYBA is based merely on fragment contributions, it can be used for the analysis of the contribution of individual molecular parts to compound synthetic accessibility. SYBA is publicly available at https://github.com/lich-uct/syba under the GNU General Public License.

Medicinal Chemistry Aware Database GDBMedChem

The generated database GDB17 enumerates 166.4 billion possible molecules up to 17 atoms of C, N, O, S and halogens following simple chemical stability and synthetic feasibility rules, however medicinal chemistry criteria are not taken into account. Here we applied rules inspired by medicinal chemistry to exclude problematic functional groups and complex molecules from GDB17, and sampled the resulting subset uniformly across molecular size, stereochemistry and polarity to form GDBMedChem as a compact collection of 10 million small molecules. This collection has reduced complexity and better synthetic accessibility than the entire GDB17 but retains higher sp³ -carbon fraction and natural product likeness scores compared to known drugs. GDBMedChem molecules are more diverse and very different from known molecules in terms of substructures and represent an unprecedented source of diversity for drug design. GDBMedChem is available for 3D-visualization, similarity searching and for download at http://gdb.unibe.ch.

Interrogating dense ligand chemical space with a forward-synthetic library

Forward-synthetic databases are an efficient way to enumerate chemical space. We explored here whether these databases are good sources of novel protein ligands and how many molecules are obtainable and in which time frame. Based on docking calculations, series of molecules were selected to gain insights into the ligand structure-activity relationship. To evaluate the novelty of compounds in a challenging way, we chose the β₂-adrenergic receptor, for which a large number of ligands is already known. Finding dissimilar ligands is thus the exception rather than the rule. Here we report on the results, the successful synthesis of 127/240 molecules in just 2 weeks, the discovery of previously unreported dissimilar ligands of the β₂-adrenergic receptor, and the optimization of one series to a K _D of 519 nM in only one round. Moreover, the finding that only 3 of 240 molecules had ever been synthesized before indicates that large parts of chemical space are unexplored.

GuacaMol: Benchmarking Models for de Novo Molecular Design

De novo design seeks to generate molecules with required property profiles by virtual design-make-test cycles. With the emergence of deep learning and neural generative models in many application areas, models for molecular design based on neural networks appeared recently and show promising results. However, the new models have not been profiled on consistent tasks, and comparative studies to well-established algorithms have only seldom been performed. To standardize the assessment of both classical and neural models for de novo molecular design, we propose an evaluation framework, GuacaMol, based on a suite of standardized benchmarks. The benchmark tasks encompass measuring the fidelity of the models to reproduce the property distribution of the training sets, the ability to generate novel molecules, the exploration and exploitation of chemical space, and a variety of single and multiobjective optimization tasks. The benchmarking open-source Python code and a leaderboard can be found on https://benevolent.ai/guacamol .

The next level in chemical space navigation: going far beyond enumerable compound libraries

Recent innovations have brought pharmacophore-driven methods for navigating virtual chemical spaces, the size of which can reach into the billions of molecules, to the fingertips of every chemist. There has been a paradigm shift in the underlying computational chemistry that drives chemical space search applications, incorporating intelligent reaction knowledge into their core so that they can readily deliver commercially available molecules as nearest neighbor hits from within giant virtual spaces. These vast resources enable medicinal chemists to execute rapid scaffold-hopping experiments, rapid hit expansion, and structure-activity relationship (SAR) exploitation in largely intellectual property (IP)-free territory and at unparalleled low cost.

Ultra-large library docking for discovering new chemotypes

Despite intense interest in expanding chemical space, libraries containing hundreds-of-millions to billions of diverse molecules have remained inaccessible. Here we investigate structure-based docking of 170 million make-on-demand compounds from 130 well-characterized reactions. The resulting library is diverse, representing over 10.7 million scaffolds that are otherwise unavailable. For each compound in the library, docking against AmpC β-lactamase (AmpC) and the D4 dopamine receptor were simulated. From the top-ranking molecules, 44 and 549 compounds were synthesized and tested for interactions with AmpC and the D4 dopamine receptor, respectively. We found a phenolate inhibitor of AmpC, which revealed a group of inhibitors without known precedent. This molecule was optimized to 77 nM, which places it among the most potent non-covalent AmpC inhibitors known. Crystal structures of this and other AmpC inhibitors confirmed the docking predictions. Against the D4 dopamine receptor, hit rates fell almost monotonically with docking score, and a hit-rate versus score curve predicted that the library contained 453,000 ligands for the D4 dopamine receptor. Of 81 new chemotypes discovered, 30 showed submicromolar activity, including a 180-pM subtype-selective agonist of the D4 dopamine receptor.

Virtual Compound Libraries in Computer-Assisted Drug Discovery

The use of virtual compound libraries in computer-assisted drug discovery has gained in popularity and has already lead to numerous successes. Here, we examine key static and dynamic virtual library concepts that have been developed over the past decade. To facilitate the search for new drugs in the vastness of chemical space, there are still several hurdles to overcome, including the current difficulties in screening and parsing efficiency and the need for more reliable vendors and accurate synthesis prediction tools. These challenges should be tackled by both the developers of virtual libraries and by their users, in order for the exploration of chemical space to live up to its potential.

QBMG: quasi-biogenic molecule generator with deep recurrent neural network

Biogenic compounds are important materials for drug discovery and chemical biology. In this work, we report a quasi-biogenic molecule generator (QBMG) to compose virtual quasi-biogenic compound libraries by means of gated recurrent unit recurrent neural networks. The library includes stereo-chemical properties, which are crucial features of natural products. QMBG can reproduce the property distribution of the underlying training set, while being able to generate realistic, novel molecules outside of the training set. Furthermore, these compounds are associated with known bioactivities. A focused compound library based on a given chemotype/scaffold can also be generated by this approach combining transfer learning technology. This approach can be used to generate virtual compound libraries for pharmaceutical lead identification and optimization.

Chemical Space Overlap with Critical Protein-Protein Interface Residues in Commercial and Specialized Small-Molecule Libraries

There is growing interest in the use of structure-based virtual screening to identify small molecules that inhibit challenging protein-protein interactions (PPIs). In this study, we investigated how effectively chemical library members docked at the PPI interface mimic the position of critical side-chain residues known as "hot spots". Three compound collections were considered, a commercially available screening collection (ChemDiv), a collection of diversity-oriented synthesis (DOS) compounds that contains natural-product-like small molecules, and a library constructed using established reactions (the "screenable chemical universe based on intuitive data organization", SCUBIDOO). Three different tight PPIs for which hot-spot residues have been identified were selected for analysis: uPAR⋅uPA, TEAD4⋅Yap1, and CaV α⋅CaV β. Analysis of library physicochemical properties was followed by docking to the PPI receptors. A pharmacophore method was used to measure overlap between small-molecule substituents and hot-spot side chains. Fragment-like conformationally restricted small molecules showed better hot-spot overlap for interfaces with well-defined pockets such as uPAR⋅uPA, whereas better overlap was observed for more complex DOS compounds in interfaces lacking a well-defined binding site such as TEAD4⋅Yap1. Virtual screening of conformationally restricted compounds targeting uPAR⋅uPA and TEAD4⋅Yap1 followed by experimental validation reinforce these findings, as the best hits were fragment-like and had few rotatable bonds for the former, while no hits were identified for the latter. Overall, such studies provide a framework for understanding PPIs in the context of additional chemical matter and new PPI definitions.

eToxPred: a machine learning-based approach to estimate the toxicity of drug candidates

BACKGROUND

The efficiency of drug development defined as a number of successfully launched new pharmaceuticals normalized by financial investments has significantly declined. Nonetheless, recent advances in high-throughput experimental techniques and computational modeling promise reductions in the costs and development times required to bring new drugs to market. The prediction of toxicity of drug candidates is one of the important components of modern drug discovery.

RESULTS

In this work, we describe eToxPred, a new approach to reliably estimate the toxicity and synthetic accessibility of small organic compounds. eToxPred employs machine learning algorithms trained on molecular fingerprints to evaluate drug candidates. The performance is assessed against multiple datasets containing known drugs, potentially hazardous chemicals, natural products, and synthetic bioactive compounds. Encouragingly, eToxPred predicts the synthetic accessibility with the mean square error of only 4% and the toxicity with the accuracy of as high as 72%.

CONCLUSIONS

eToxPred can be incorporated into protocols to construct custom libraries for virtual screening in order to filter out those drug candidates that are potentially toxic or would be difficult to synthesize. It is freely available as a stand-alone software at https://github.com/pulimeng/etoxpred .

Virtual Chemical Libraries

Advances in computer processing speed and storage capacity have enabled researchers to generate virtual chemical libraries containing billions of molecules. While these numbers appear large, they are only a small fraction of the number of organic molecules that could potentially be synthesized. This review provides an overview of recent advances in the generation and use of virtual chemical libraries in medicinal chemistry. We also consider the practical implications of these libraries in drug discovery programs and highlight a number of current and future challenges.

Selectivity Challenges in Docking Screens for GPCR Targets and Antitargets

To investigate large library docking's ability to find molecules with joint activity against on-targets and selectivity versus antitargets, the dopamine D₂ and serotonin 5-HT_2A receptors were targeted, seeking selectivity against the histamine H₁ receptor. In a second campaign, κ-opioid receptor ligands were sought with selectivity versus the μ-opioid receptor. While hit rates ranged from 40% to 63% against the on-targets, they were just as good against the antitargets, even though the molecules were selected for their putative lack of binding to the off-targets. Affinities, too, were often as good or better for the off-targets. Even though it was occasionally possible to find selective molecules, such as a mid-nanomolar D₂/5-HT_2A ligand with 21-fold selectivity versus the H₁ receptor, this was the exception. Whereas false-negatives are tolerable in docking screens against on-targets, they are intolerable against antitargets; addressing this problem may demand new strategies in the field.

Integrated Strategy for Lead Optimization Based on Fragment Growing: The Diversity-Oriented-Target-Focused-Synthesis Approach

Over the past few decades, hit identification has been greatly facilitated by advances in high-throughput and fragment-based screenings. One major hurdle remaining in drug discovery is process automation of hit-to-lead (H2L) optimization. Here, we report a time- and cost-efficient integrated strategy for H2L optimization as well as a partially automated design of potent chemical probes consisting of a focused-chemical-library design and virtual screening coupled with robotic diversity-oriented de novo synthesis and automated in vitro evaluation. The virtual library is generated by combining an activated fragment, corresponding to the substructure binding to the target, with a collection of functionalized building blocks using in silico encoded chemical reactions carefully chosen from a list of one-step organic transformations relevant in medicinal chemistry. The proof of concept was demonstrated using the optimization of bromodomain inhibitors as a test case, leading to the validation of several compounds with improved affinity by several orders of magnitude.

Chemical Space Expansion of Bromodomain Ligands Guided by in Silico Virtual Couplings (AutoCouple)

Expanding the chemical space and simultaneously ensuring synthetic accessibility is of upmost importance, not only for the discovery of effective binders for novel protein classes but, more importantly, for the development of compounds against hard-to-drug proteins. Here, we present AutoCouple, a de novo approach to computational ligand design focused on the diversity-oriented generation of chemical entities via virtual couplings. In a benchmark application, chemically diverse compounds with low-nanomolar potency for the CBP bromodomain and high selectivity against the BRD4(1) bromodomain were achieved by the synthesis of about 50 derivatives of the original fragment. The binding mode was confirmed by X-ray crystallography, target engagement in cells was demonstrated, and antiproliferative activity was showcased in three cancer cell lines. These results reveal AutoCouple as a useful in silico coupling method to expand the chemical space in hit optimization campaigns resulting in potent, selective, and cell permeable bromodomain ligands.