Search

Many therapeutic drugs are compounds that can be represented by simple chemical structures, which contain important determinants of affinity at the site of action. Recently, graph convolutional neural network (GCN) models have exhibited excellent results in classifying the activity of such compounds. For models that make quantitative predictions of activity, more complex information has been utilized, such as the three-dimensional structures of compounds and the amino acid sequences of their respective target proteins. As another approach, we hypothesized that if sufficient experimental data were available and there were enough nodes in hidden layers, a simple compound representation would quantitatively predict activity with satisfactory accuracy. In this study, we report that GCN models constructed solely from the two-dimensional structural information of compounds demonstrated a high degree of activity predictability against 127 diverse targets from the ChEMBL database. Using the information entropy as a metric, we also show that the structural diversity had less effect on the prediction performance. Finally, we report that virtual screening using the constructed model identified a new serotonin transporter inhibitor with activity comparable to that of a marketed drug in vitro and exhibited antidepressant effects in behavioural studies., 化合物の薬理作用を予測する技術を開発 --薬理作用ビッグデータを用いて--. 京都大学プレスリリース. 2021-01-13.

With the public availability of large data sources such as ChEMBLdb and the Open PHACTS Discovery Platform, retrieval of data sets for certain protein targets of interest with consistent assay conditions is no longer a time consuming process. Especially the use of workflow engines such as KNIME or Pipeline Pilot allows complex queries and enables to simultaneously search for several targets. Data can then directly be used as input to various ligand- and structure-based studies. In this contribution, using in-house projects on P-gp inhibition, transporter selectivity, and TRPV1 modulation we outline how the incorporation of linked life science data in the daily execution of projects allowed to expand our approaches from conventional Hansch analysis to complex, integrated multilayer models. Electronic supplementary material The online version of this article (doi:10.1007/s10822-016-9990-4) contains supplementary material, which is available to authorized users.

The emergence of a number of publicly available bioactivity databases, such as ChEMBL, PubChem BioAssay and BindingDB, has raised awareness about the topics of data curation, quality and integrity. Here we provide an overview and discussion of the current and future approaches to activity, assay and target data curation of the ChEMBL database. This curation process involves several manual and automated steps and aims to: (1) maximise data accessibility and comparability; (2) improve data integrity and flag outliers, ambiguities and potential errors; and (3) add further curated annotations and mappings thus increasing the usefulness and accuracy of the ChEMBL data for all users and modellers in particular. Issues related to activity, assay and target data curation and integrity along with their potential impact for users of the data are discussed, alongside robust selection and filter strategies in order to avoid or minimise these, depending on the desired application.

Background Small molecules are information carriers that make cells aware of external changes and couple internal metabolic and signalling pathway systems with each other. In some specific physiological status, natural or artificial molecules are used to interact with selective biological targets to activate or inhibit their functions to achieve expected biological and physiological output. Millions of years of evolution have optimized biological processes and pathways and now the endocrine and immune system cannot work properly without some key small molecules. In the past thousands of years, the human race has managed to find many medicines against diseases by trail-and-error experience. In the recent decades, with the deepening understanding of life and the progress of molecular biology, researchers spare no effort to design molecules targeting one or two key enzymes and receptors related to corresponding diseases. But recent studies in pharmacogenomics have shown that polypharmacology may be necessary for the effects of drugs, which challenge the paradigm, ‘one drug, one target, one disease’. Nowadays, cheminformatics and structural biology can help us reasonably take advantage of the polypharmacology to design next-generation promiscuous drugs and drug combination therapies. Results 234,591 protein–ligand interactions were extracted from ChEMBL. By the 2D structure similarity, 13,769 ligand emerged from 156,151 distinct ligands which were recognized by 1477 proteins. Ligand cluster- and sequence-based protein networks (LCBN, SBN) were constructed, compared and analysed. For assisting compound designing, exploring polypharmacology and finding possible drug combination, we integrated the pathway, disease, drug adverse reaction and the relationship of targets and ligand clusters into the web platform, ePlatton, which is available at http://www.megabionet.org/eplatton. Conclusions Although there were some disagreements between the LCBN and SBN, communities in both networks were largely the same with normalized mutual information at 0.9. The study of target and ligand cluster promiscuity underlying the LCBN showed that light ligand clusters were more promiscuous than the heavy one and that highly connected nodes tended to be protein kinases and involved in phosphorylation. ePlatton considerably reduced the redundancy of the ligand set of targets and made it easy to deduce the possible relationship between compounds and targets, pathways and side effects. ePlatton behaved reliably in validation experiments and also fast in virtual screening and information retrieval.Graphical abstractCluster exemplars and ePlatton’s mechanism. Electronic supplementary material The online version of this article (doi:10.1186/s13321-016-0135-5) contains supplementary material, which is available to authorized users.

Background The concept of molecular similarity is one of the central ideas in cheminformatics, despite the fact that it is ill-defined and rather difficult to assess objectively. Here we propose a practical definition of molecular similarity in the context of drug discovery: molecules A and B are similar if a medicinal chemist would be likely to synthesise and test them around the same time as part of the same medicinal chemistry program. The attraction of such a definition is that it matches one of the key uses of similarity measures in early-stage drug discovery. If we make the assumption that molecules in the same compound activity table in a medicinal chemistry paper were considered similar by the authors of the paper, we can create a dataset of similar molecules from the medicinal chemistry literature. Furthermore, molecules with decreasing levels of similarity to a reference can be found by either ordering molecules in an activity table by their activity, or by considering activity tables in different papers which have at least one molecule in common. Results Using this procedure with activity data from ChEMBL, we have created two benchmark datasets for structural similarity that can be used to guide the development of improved measures. Compared to similar results from a virtual screen, these benchmarks are an order of magnitude more sensitive to differences between fingerprints both because of their size and because they avoid loss of statistical power due to the use of mean scores or ranks. We measure the performance of 28 different fingerprints on the benchmark sets and compare the results to those from the Riniker and Landrum (J Cheminf 5:26, 2013. doi:10.1186/1758-2946-5-26) ligand-based virtual screening benchmark. Conclusions Extended-connectivity fingerprints of diameter 4 and 6 are among the best performing fingerprints when ranking diverse structures by similarity, as is the topological torsion fingerprint. However, when ranking very close analogues, the atom pair fingerprint outperforms the others tested. When ranking diverse structures or carrying out a virtual screen, we find that the performance of the ECFP fingerprints significantly improves if the bit-vector length is increased from 1024 to 16,384.Graphical abstractAn example series from one of the benchmark datasets. Each fingerprint is assessed on its ability to reproduce a specific series order. Electronic supplementary material The online version of this article (doi:10.1186/s13321-016-0148-0) contains supplementary material, which is available to authorized users.

The process of discovering new drugs is a lengthy, time-consuming and expensive process. Modern day drug discovery relies heavily on the rapid identification of novel ‘targets’, usually proteins that can be modulated by small molecule drugs to cure or minimise the effects of a disease. Of the 20,000 proteins currently reported as comprising the human proteome, just under a quarter of these can potentially be modulated by known small molecules Storing information in curated, actively maintained drug discovery databases can help researchers access current drug discovery information quickly. However with the increase in the amount of data generated from both experimental and in silico efforts, databases can become very large very quickly and information retrieval from them can become a challenge. The development of database tools that facilitate rapid information retrieval is important to keep up with the growth of databases. We have developed a Gene Ontology-based navigation tool (Gene Ontology Tree) to help users retrieve biological information to single protein targets in the ChEMBL drug discovery database. 99 % of single protein targets in ChEMBL have at least one GO annotation associated with them. There are 12,500 GO terms associated to 6200 protein targets in the ChEMBL database resulting in a total of 140,000 annotations. The slim we have created, the ‘ChEMBL protein target slim’ allows broad categorisation of the biology of 90 % of the protein targets using just 300 high level, informative GO terms. We used the GO slim method of assigning fewer higher level GO groupings to numerous very specific lower level terms derived from the GOA to describe a set of GO terms relevant to proteins in ChEMBL. We then used the slim created to provide a web based tool that allows a quick and easy navigation of protein target space. Terms from the GO are used to capture information on protein molecular function, biological process and subcellular localisations. The ChEMBL database also provides compound information for small molecules that have been tested for their effects on these protein targets. The ‘ChEMBL protein target slim’ provides a means of firstly describing the biology of protein drug targets and secondly allows users to easily establish a connection between biological and chemical information regarding drugs and drug targets in ChEMBL. The ‘ChEMBL protein target slim’ is available as a browsable ‘Gene Ontology Tree’ on the ChEMBL site under the browse targets tab ( https://www.ebi.ac.uk/chembl/target/browser ). A ChEMBL protein target slim OBO file containing the GO slim terms pertinent to ChEMBL is available from the GOC website ( http://geneontology.org/page/go-slim-and-subset-guide ). We have created a protein target navigation tool based on the ‘ChEMBL protein target slim’. The ‘ChEMBL protein target slim’ provides a way of browsing protein targets in ChEMBL using high level GO terms that describe the molecular functions, processes and subcellular localisations of protein drug targets in drug discovery. The tool also allows user to establish a link between ontological groupings representing protein target biology to relevant compound information in ChEMBL. We have demonstrated by the use of a simple example how the ‘ChEMBL protein target slim’ can be used to link biological processes with drug information based on the information in the ChEMBL database. The tool has potential to aid in areas of drug discovery such as drug repurposing studies or drug-disease-protein pathways.

Background: It is now widely recognized that there is an urgent need for new antibacterial drugs, with novel mechanisms of action, to combat the rise of multi-drug resistant bacteria. However, few new compounds are reaching the market. Antibacterial drug discovery projects often succeed in identifying potent molecules in biochemical assays but have been beset by difficulties in obtaining antibacterial activity. A commonly held view, based on analysis of marketed antibacterial compounds, is that antibacterial drugs possess very different physicochemical properties to other drugs, and that this profile is required for antibacterial activity. Results: We have re-examined this issue by performing a cheminformatics analysis of the literature data available in the ChEMBL database. The physicochemical properties of compounds with a recorded activity in an antibacterial assay were calculated and compared to two other datasets extracted from ChEMBL, marketed antibacterials and drugs marketed for other therapeutic indications. The chemical class of the compounds and Gram-negative/Gram-positive profile were also investigated. This analysis shows that compounds with antibacterial activity have physicochemical property profiles very similar to other drug classes. Conclusions: The observation that many current antibacterial drugs lie in regions of physicochemical property space far from conventional small molecule therapeutics is correct. However, the inference that a compound must lie in one of these "outlier" regions in order to possess antibacterial activity is not supported by our analysis., peer-reviewed

Background The Plasmodium falciparum M18 Aspartyl Aminopeptidase (PfM18AAP) is only aspartyl aminopeptidase which is found in the genome of P. falciparum and is essential for its survival. The PfM18AAP enzyme performs various functions in the parasite and the erythrocytic host such as hemoglobin digestion, erythrocyte invasion, parasite growth and parasite escape from the host cell. It is a valid target to develop antimalarial drugs. In the present work, we employed 3D QSAR modeling, pharmacophore modeling, and molecular docking to identify novel potent inhibitors that bind with M18AAP of P. falciparum. Results The PLSR QSAR model showed highest value for correlation coefficient r2 (88 %) and predictive correlation coefficient (pred_r2) =0.6101 for external test set among all QSAR models. The pharmacophore modeling identified DHRR (one hydrogen donor, one hydrophobic group, and two aromatic rings) as an essential feature of PfM18AAP inhibitors. The combined approach of 3D QSAR, pharmacophore, and structure-based molecular docking yielded 10 novel PfM18AAP inhibitors from ChEMBL antimalarial library, 2 novel inhibitors from each derivative of quinine, chloroquine, 8-aminoquinoline and 10 novel inhibitors from WHO antimalarial drugs. Additionally, high throughput virtual screening identified top 10 compounds as antimalarial leads showing G-scores -12.50 to -10.45 (in kcal/mol), compared with control compounds(G-scores -7.80 to -4.70) which are known antimalarial M18AAP inhibitors (AID743024). This result indicates these novel compounds have the best binding affinity for PfM18AAP. Conclusion The 3D QSAR models of PfM18AAP inhibitors provided useful information about the structural characteristics of inhibitors which are contributors of the inhibitory potency. Interestingly, In this studies, we extrapolate that the derivatives of quinine, chloroquine, and 8-aminoquinoline, for which there is no specific target has been identified till date, might show the antimalarial effect by interacting with PfM18AAP. Electronic supplementary material The online version of this article (doi:10.1186/s12900-016-0063-7) contains supplementary material, which is available to authorized users.

The analysis of the biological effects of small molecules by mining lot of growing databases is an important task in the field of pharmaceutical sciences. To identify potential new similar drugs or to assess health risks from chemicals requires detailed knowledge about compounds. The pursued project is aiming at the integration of existing data resources of small molecules that are combined with tools for the prediction of biological effects based on molecular interactions. Users can query the system for a given compound and retrieve information on questions such as: • Which other similar compounds exist? • Which proteins are mentioned in the literature in the same context of the given or similar compound(s)? • Has the compound been tested in a bioassay? • Is the compound toxic? • Is the compound purchasable? • Is the compound patented? Answers to the above questions will be given by the web interface dynamically from the web server. The system further records the users’ choices and learns from them such that more important information shown more prominently. The system is based on a large compound library consisting of above 70 million molecules collected from different publicly available resources (e.g., PubChem [1], ChEMBL [2], ZINC [3], StreptomeDB [4]) and will have the ability to accommodate new databases easily. An integrated text mining tool (CIL [5]) provides information on functional relations of queried compounds to proteins and refers to articles describing them. Tools for the prediction of biological effects such as SuperPred [6] will also be included. Subsequently, the results will be adapted to the requirements of the user and gives an extensive digest of relevant information for pharmaceutical researchers.

Cheminformatic methods, such as Virtual Screening, constitute a vital part of modern drug design process. This technique enables not only viable prediction of physicochemical properties of the molecules, but also effective database mining, being particularly useful tool in search for ligands of desired activity. Successful performance in case of single target drugs, implies a potential to extend its capabilities to compounds bearing desired activity towards multiple receptors. In this research, we present application of Structural Interaction Fingerprints (SIFts) [1] combined with Machine Learning (ML) as a method to select single- and multi-target ligands from the docking results. A handful of protein kinases pairs was designated as targets. Collection of pseudo selective compounds, with various activity profiles, was acquired from ChEMBL database. Furthermore, for each target a set of ligands of various activity was aggregated. Decoy structures were random ligands from ZINC database. The compounds were docked into respective proteins, and SIFts were calculated for each protein-ligand complex. Training sets used in ML experiments consisted of cluster centroids of active and inactive ligands, whereas test sets were composed of remaining compounds, that returned docking poses. The key aim of this study is to develop a viable method to filter the docking results, so that the compounds meeting desired activity profile are selected.

Background: The World Anti-Doping Agency (WADA) publishes the Prohibited List, a manually compiled international standard of substances and methods prohibited in-competition, out-of-competition and in particular sports. It would be ideal to be able to identify all substances that have one or more performance-enhancing pharmacological actions in an automated, fast and cost effective way. Here, we use experimental data derived from the ChEMBL database (~7,000,000 activity records for 1,300,000 compounds) to build a database model that takes into account both structure and experimental information, and use this database to predict both on-target and off-target interactions between these molecules and targets relevant to doping in sport. Results: The ChEMBL database was screened and eight well populated categories of activities (Ki, Kd, EC50, ED50, activity, potency, inhibition and IC50) were used for a rule-based filtering process to define the labels “active” or “inactive”. The “active” compounds for each of the ChEMBL families were thereby defined and these populated our bioactivity-based filtered families. A structure-based clustering step was subsequently performed in order to split families with more than one distinct chemical scaffold. This produced refined families, whose members share both a common chemical scaffold and bioactivity against a common target in ChEMBL. Conclusions: We have used the Parzen-Rosenblatt machine learning approach to test whether compounds in ChEMBL can be correctly predicted to belong to their appropriate refined families. Validation tests using the refined families gave a significant increase in predictivity compared with the filtered or with the original families. Out of 61,660 queries in our Monte Carlo cross-validation, belonging to 19,639 refined families, 41,300 (66.98%) had the parent family as the top prediction and 53,797 (87.25%) had the parent family in the top four hits. Having thus validated our approach, we used it to identify the protein targets associated with the WADA prohibited classes. For compounds where we do not have experimental data, we use their computed patterns of interaction with protein targets to make predictions of bioactivity. We hope that other groups will test these predictions experimentally in the future. Publisher PDF

Nowadays supervised classification, based on the concept of pattern recognition, is an integral part of virtual screening. The central idea of supervised classification in chemoinformatics is to design a classifying algorithm that accurately assigns a new molecule to one of a set of predefined classes. Naturally, probabilistic classifiers can be far more useful than hard point classifiers in making a decision on problems [1], such as virtual screening, where there is an associated risk in classifying an instance to one class or the other. For their conceptual simplicity and computational efficiency probabilistic classification methods based on the Naive Bayes concept are widely employed in chemoinformatics. The simplicity of the Naive Bayes is due to the assumption that the descriptors representing the molecule one desires to classify are statistically independent. Unfortunately it is well documented that when the molecular descriptors are binary-valued - which is often the case in chemoinformatics - and thus take values of 0 or 1 the Naive Bayesian classifier can only act as a linear classifier in the descriptor space. Techniques such as the Parzen-Window approach can address the above shortcomings but suffer from being computationally expensive as they require one to retain all the training dataset in core memory [2,3]. In an attempt to address the above mentioned drawbacks, a new probabilistic classifier is proposed which uses randomized sub-sampling of the descriptor space. The proposed algorithm generates better class membership predictions than its Naive Bayesian counterpart on classifying molecules that are non-linearly separable in descriptor space. We present a realistic test of the new method by classifying large chemical datasets generated from the ChEMBL database [4].

Background The current chemical space of known small molecules is estimated to exceed 1060 structures. Though the largest physical compound repositories contain only a few tens of millions of unique compounds, virtual screening of databases of this size is still difficult. In recent years, the application of physicochemical descriptor-based profiling, such as Lipinski's rule-of-five for drug-likeness and Oprea's criteria of lead-likeness, as early stage filters in drug discovery has gained widespread acceptance. In the current study, we outline a kinase-likeness scoring function based on known kinase inhibitors. Results The method employs a collection of 22,615 known kinase inhibitors from the ChEMBL database. A kinase-likeness score is computed using statistical analysis of nine key physicochemical descriptors for these inhibitors. Based on this score, the kinase-likeness of four publicly and commercially available databases, i.e., National Cancer Institute database (NCI), the Natural Products database (NPD), the National Institute of Health's Molecular Libraries Small Molecule Repository (MLSMR), and the World Drug Index (WDI) database, is analyzed. Three of these databases, i.e., NCI, NPD, and MLSMR are frequently used in the virtual screening of kinase inhibitors, while the fourth WDI database is for comparison since it covers a wide range of known chemical space. Based on the kinase-likeness score, a kinase-focused library is also developed and tested against three different kinase targets selected from three different branches of the human kinome tree. Conclusions Our proposed methodology is one of the first that explores how the narrow chemical space of kinase inhibitors and its relevant physicochemical information can be utilized to build kinase-focused libraries and prioritize pre-existing compound databases for screening. We have shown that focused libraries generated by filtering compounds using the kinase-likeness score have, on average, better docking scores than an equivalent number of randomly selected compounds. Beyond library design, our findings also impact the broader efforts to identify kinase inhibitors by screening pre-existing compound libraries. Currently, the NCI library is the most commonly used database for screening kinase inhibitors. Our research suggests that other libraries, such as MLSMR, are more kinase-like and should be given priority in kinase screenings.

Background There is a growing body of literature describing the properties of marketed drugs, the concept of drug-likeness and the vastness of chemical space. In that context, enumerative combinatorics with simple atomic components may be useful in the conception and design of structurally novel compounds for expanding and enhancing high-throughput screening (HTS) libraries. Results A random combination of mono- and diatomic carbon, hydrogen, nitrogen, and oxygen containing components in the absence of molecular weight constraints but with the ability to form rings affords virtual compounds that fall in bulk physicochemical space typically associated with drugs, but whose ring assemblies fall in new or under-represented areas of chemical shape space. When compared against compounds in the ChEMBL_14, MDDR, Drug Bank and Dictionary of Natural Products, the percentage of virtual compounds with a Tanimoto index of 1.0 (ECFP_4) was found to be as high as 0.21. Depending on therapeutic target, this value may be in range of what might be expected from an experimental HTS campaign in terms of a true hit rate. Conclusion Virtual compounds derived through enumerative combinatorics of simple atomic components have drug-like properties with ring assemblies that fall in new or under-represented areas of shape space. Structures derived in this manner could provide the starting point or inspiration for the design of structurally novel scaffolds in an unbiased fashion.