Public Domain HTS Fingerprints: Design and Evaluation of Compound Bioactivity Profiles from PubChem's Bioassay Repository.

Molecular profiling efforts aim at characterizing the biological actions of small molecules by screening them in hundreds of different biochemical and/or cell-based assays. Together, these assays yield a rich data landscape of target-based and phenotypic effects of the tested compounds. However, submitting an entire compound library to a molecular profiling panel can easily become cost-prohibitive. Here, we make use of historical screening assays to create comprehensive bioactivity profiles for more than 300 000 small molecules. These bioactivity profiles, termed PubChem high-throughput screening fingerprints (PubChem HTSFPs), report small molecule activities in 243 different PubChem bioassays. Although the assays originate from originally independently pursued drug or probe discovery projects, we demonstrate their value as molecular signatures when used in combination. We use these PubChem HTSFPs as molecular descriptors in hit expansion experiments for 33 different targets and phenotypes, showing that, on average, they lead to 27 times as many hits in a set of 1000 chosen molecules as a random screening subset of the same size (average ROC score: 0.82). Moreover, we demonstrate that PubChem HTSFPs retrieve hits that are structurally diverse and distinct from active compounds retrieved by chemical similarity-based hit expansion methods. PubChem HTSFPs are made freely available for the chemical biology research community.

Identifying mechanism-of-action targets for drugs and probes.

Notwithstanding their key roles in therapy and as biological probes, 7% of approved drugs are purported to have no known primary target, and up to 18% lack a well-defined mechanism of action. Using a chemoinformatics approach, we sought to "de-orphanize" drugs that lack primary targets. Surprisingly, targets could be easily predicted for many: Whereas these targets were not known to us nor to the common databases, most could be confirmed by literature search, leaving only 13 Food and Drug Administration-approved drugs with unknown targets; the number of drugs without molecular targets likely is far fewer than reported. The number of worldwide drugs without reasonable molecular targets similarly dropped, from 352 (25%) to 44 (4%). Nevertheless, there remained at least seven drugs for which reasonable mechanism-of-action targets were unknown but could be predicted, including the antitussives clemastine, cloperastine, and nepinalone; the antiemetic benzquinamide; the muscle relaxant cyclobenzaprine; the analgesic nefopam; and the immunomodulator lobenzarit. For each, predicted targets were confirmed experimentally, with affinities within their physiological concentration ranges. Turning this question on its head, we next asked which drugs were specific enough to act as chemical probes. Over 100 drugs met the standard criteria for probes, and 40 did so by more stringent criteria. A chemical information approach to drug-target association can guide therapeutic development and reveal applications to probe biology, a focus of much current interest.

Indexing molecules with chemical graph identifiers.

Fast and robust algorithms for indexing molecules have been historically considered strategic tools for the management and storage of large chemical libraries. This work introduces a modified and further extended version of the molecular equivalence number naming adaptation of the Morgan algorithm (J Chem Inf Comput Sci 2001, 41, 181-185) for the generation of a chemical graph identifier (CGI). This new version corrects for the collisions recognized in the original adaptation and includes the ability to deal with graph canonicalization, ensembles (salts), and isomerism (tautomerism, regioisomerism, optical isomerism, and geometrical isomerism) in a flexible manner. Validation of the current CGI implementation was performed on the open NCI database and the drug-like subset of the ZINC database containing 260,071 and 5,348,089 structures, respectively. The results were compared with those obtained with some of the most widely used indexing codes, such as the CACTVS hash code and the new InChIKey. The analyses emphasize the fact that compound management activities, like duplicate analysis of chemical libraries, are sensitive to the exact definition of compound uniqueness and thus still depend, to a minor extent, on the type and flexibility of the molecular index being used.

In silico directed chemical probing of the adenosine receptor family.

One of the grand challenges in chemical biology is identifying a small-molecule modulator for each individual function of all human proteins. Instead of targeting one protein at a time, an efficient approach to address this challenge is to target entire protein families by taking advantage of the relatively high levels of chemical promiscuity observed within certain boundaries of sequence phylogeny. We recently developed a computational approach to identifying the potential protein targets of compounds based on their similarity to known bioactive molecules for almost 700 targets. Here, we describe the direct identification of novel antagonists for all four adenosine receptor subtypes by applying our virtual profiling approach to a unique synthesis-driven chemical collection composed of 482 biologically-orphan molecules. These results illustrate the potential role of in silico target profiling to guide efficiently screening campaigns directed to discover new chemical probes for all members of a protein family.

Coverage and bias in chemical library design.

The design of chemical libraries directed to target classes is an activity that requires the availability of ligand pharmacological data and/or protein structural data. On the basis of the knowledge derived from these data, chemical libraries directed mainly to G protein-coupled receptors, kinases, proteases, and nuclear receptors have been assembled. However, current design strategies widely overlook assessing the potential ability of the compounds contained in a focused library to provide uniform ample coverage of the protein family they intend to target. Here, we discuss the use of in silico target profiling methods as a means to estimate the actual scope of chemical libraries to probe entire protein families and illustrate its applicability in optimizing the composition of compound sets to achieve maximum coverage of the family with minimum bias to particular targets.

A ligand-based approach to mining the chemogenomic space of drugs.

The practical implementation and validation of a ligand-based approach to mining the chemogenomic space of drugs is presented and applied to the in silico target profiling of 767 drugs against 684 targets of therapeutic relevance. The results reveal that drugs targeting aminergic G protein-coupled receptors (GPCRs) show the most promiscuous pharmacological profiles. The detection of cross-pharmacologies between aminergic GPCRs and the opioid, sigma, NMDA, and 5-HT3 receptors aggravate the potential promiscuity of those drugs, predominantly including analgesics, antidepressants, and antipsychotics.

Docking and Linking of Fragments To Discover Jumonji Histone Demethylase Inhibitors.

Development of tool molecules that inhibit Jumonji demethylases allows for the investigation of cancer-associated transcription. While scaffolds such as 2,4-pyridinedicarboxylic acid (2,4-PDCA) are potent inhibitors, they exhibit limited selectivity. To discover new inhibitors for the KDM4 demethylases, enzymes overexpressed in several cancers, we docked a library of 600,000 fragments into the high-resolution structure of KDM4A. Among the most interesting chemotypes were the 5-aminosalicylates, which docked in two distinct but overlapping orientations. Docking poses informed the design of covalently linked fragment compounds, which were further derivatized. This combined approach improved affinity by ∼ 3 log-orders to yield compound 35 (Ki = 43 nM). Several hybrid inhibitors were selective for KDM4C over the related enzymes FIH, KDM2A, and KDM6B while lacking selectivity against the KDM3 and KDM5 subfamilies. Cocrystal structures corroborated the docking predictions. This study extends the use of structure-based docking from fragment discovery to fragment linking optimization, yielding novel KDM4 inhibitors.

Identification of small molecule inhibitors of amyloid ß-induced neuronal apoptosis acting through the imidazoline I(2) receptor.

Aberrant activation of signaling pathways plays a pivotal role in central nervous system disorders, such as Alzheimer's disease (AD). Using a combination of virtual screening and experimental testing, novel small molecule inhibitors of tPA-mediated extracellular signal-regulated kinase (Erk)1/2 activation were identified that provide higher levels of neuroprotection from Aß-induced apoptosis than Memantine, the most recently FDA-approved drug for AD treatment. Subsequent target deconvolution efforts revealed that they all share low micromolar affinity for the imidazoline I(2) receptor, while being devoid of any significant affinity to a list of AD-relevant targets, including the N-methyl-d-aspartate receptor (NMDAR), acetylcholinesterase (AChE), and monoamine oxidase B (MAO-B). Targeting the imidazoline I(2) receptor emerges as a new mechanism of action to inhibit tPA-induced signaling in neurons for the treatment of AD and other neurodegenerative diseases.

A chemical screen identifies class a g-protein coupled receptors as regulators of cilia.

Normal cilia length and motility are critical for proper cellular function. Prior studies of the regulation of ciliary structure and length have primarily focused on the intraflagellar transport machinery and motor proteins required for ciliary assembly and disassembly. However, several mutants with abnormal length flagella highlight the importance of signaling proteins as well. In this study, an unbiased chemical screen was performed to uncover signaling pathways that are critical for ciliogenesis and length regulation using flagella of the green alga Chlamydomonas reinhardtii as a model. The annotated Sigma LOPAC1280 chemical library was screened for effects on flagellar length, motility, and severing as well as cell viability. Assay data were clustered to identify pathways regulating flagella. The most frequent target found to be involved in flagellar length regulation was the family of dopamine binding G-protein coupled receptors (GPCRs). In mammalian cells, cilium length could indeed be altered with expression of the dopamine D1 receptor. Our screen thus reveals signaling pathways that are potentially critical for ciliary formation, resorption, and length maintenance, which represent candidate targets for therapeutic intervention of disorders involving ciliary malformation and malfunction.

New chromene scaffolds for adenosine A(2A) receptors: synthesis, pharmacology and structure-activity relationships.

In silico screening of a collection of 1584 academic compounds identified a small molecule hit for the human adenosine A(2A) receptor (pK(i) = 6.2) containing a novel chromene scaffold (3a). To explore the structure-activity relationships of this new chemical series for adenosine receptors, a focused library of 43 2H-chromene-3-carboxamide derivatives was synthesized and tested in radioligand binding assays at human adenosine A(1), A(2A), A(2B) and A(3) receptors. The series was found to be enriched with bioactive compounds for adenosine receptors, with 14 molecules showing submicromolar affinity (pK(i) ≥ 6.0) for at least one adenosine receptor subtype. These results provide evidence that the chromene scaffold, a core structure present in natural products from a wide variety of plants, vegetables, and fruits, constitutes a valuable source for novel therapeutic agents.

Conciliating binding efficiency and polypharmacology.

The association between molecular size and risk of failure has promoted the use of binding efficiency as a prioritization metric in lead selection. Even though by extension it is often referred to as "ligand efficiency", the concept was originally conceived to be strictly applicable to comparing the binding efficiencies of ligands for a single target. With current trends in designing drugs to bind efficiently to multiple targets, a revision of the original binding efficiency definition is carried out. To this aim, the dependency of binding efficiency on polypharmacology is highlighted in a retrospective analysis of a set of antipsychotic drugs. Statistical standardization of target binding efficiencies relative to basal values obtained from a large background of medicinal chemistry compounds is proposed as a means to conciliate the concepts of binding efficiency and polypharmacology. Finally, the interplay between binding efficiency and therapeutic efficacy for optimizing natural products, random hits, and fragments is discussed.

The topology of drug-target interaction networks: implicit dependence on drug properties and target families.

The availability of interaction data between small molecule drugs and protein targets has increased substantially in recent years. Using seven different databases, we were able to assemble a total of 4767 unique interactions between 802 drugs and 480 targets, which means that on average every drug is currently acknowledged to interact with 6 targets. The application of network theory to the analysis of these data reveals an unexpectedly complex picture of drug-target interactions. The results confirm that the topology of drug-target networks depends implicitly on data completeness, drug properties, and target families. The implications for drug discovery are discussed.

SHED: Shannon entropy descriptors from topological feature distributions.

A novel set of molecular descriptors called SHED (SHannon Entropy Descriptors) is presented. They are derived from distributions of atom-centered feature pairs extracted directly from the topology of molecules. The value of a SHED is then obtained by applying the information-theoretical concept of Shannon entropy to quantify the variability in a feature-pair distribution. The collection of SHED values reflecting the overall distribution of pharmacophoric features in a molecule constitutes its SHED profile. Similarity between pairs of molecules is then assessed by calculating the Euclidean distance of their SHED profiles. Under the assumption that molecules having similar pharmacological profiles should contain similar features distributed in a similar manner, examples are given to show the ability of SHED for scaffold hopping in virtual chemical screening and pharmacological profiling compared to that of substructural BCI fingerprints and three-dimensional GRIND descriptors.

Ligand-based approach to in silico pharmacology: nuclear receptor profiling.

Bioactive ligands are a valuable and increasingly accessible source of information about protein targets. On the basis of this statement, a list of 25 nuclear receptors was described by a series of bioactive ligands extracted directly from bibliographical sources, stored properly in an annotated chemical library, and mathematically represented using the recently reported SHED molecular descriptors. Analysis of this ligand information allowed for derivation of a threshold of nuclear receptor concern. If the similarity of one molecule to any of the molecules annotated to one particular nuclear receptor is below that threshold, the molecule receives an alert on the probability of having affinity below 10 microM for that nuclear receptor. On this basis, a linkage map was constructed that reveals the interaction network of nuclear receptors from the perspective of their active ligands. This ligand-based approach to nuclear receptor profiling was subsequently applied to four external chemical libraries of 10,000 molecules targeted to proteases, kinases, ion channels, and G protein-coupled receptors. The percentage of each library that returned an alert on at least one nuclear receptor was reasonably low and varied between 4.4 and 9.7%. In addition, ligand-based nuclear receptor profiling of a set of 2944 drugs provided an alert for 153 drugs. For some of them, namely, acitretin, telmisartan, phenyltoloxamine, tazarotene, and flumazenil, bibliographical evidence could be found indicating that those drugs may indeed have some potential off-target residual affinity for the nuclear receptors annotated. Overall, the present findings suggest that ligand-based approaches to protein family profiling appear as a promising means toward the establishment of novel tools for in silico pharmacology.