A quantitative high-throughput screen identifies compounds that lower expression of the SCA2-and ALS-associated gene ATXN2.

CAG repeat expansions in the ATXN2 (ataxin-2) gene can cause the autosomal dominant disorder spinocerebellar ataxia type 2 (SCA2) as well as increase the risk of ALS. Abnormal molecular, motor, and neurophysiological phenotypes in SCA2 mouse models are normalized by lowering ATXN2 transcription, and reduction of nonmutant Atxn2 expression has been shown to increase the life span of mice overexpressing the TDP-43 (transactive response DNA-binding protein 43 kDa) ALS protein, demonstrating the potential benefits of targeting ATXN2 transcription in humans. Here, we describe a quantitative high-throughput screen to identify compounds that lower ATXN2 transcription. We screened 428,759 compounds in a multiplexed assay using an ATXN2-luciferase reporter in human embryonic kidney 293 (HEK-293) cells and identified a diverse set of compounds capable of lowering ATXN2 transcription. We observed dose-dependent reductions of endogenous ATXN2 in HEK-293 cells treated with procillaridin A, 17-dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG), and heat shock protein 990 (HSP990), known inhibitors of HSP90 and Na+/K+-ATPases. Furthermore, HEK-293 cells expressing polyglutamine-expanded ATXN2-Q58 treated with 17-DMAG had minimally detectable ATXN2, as well as normalized markers of autophagy and endoplasmic reticulum stress, including STAU1 (Staufen 1), molecular target of rapamycin, p62, LC3-II (microtubule-associated protein 1A/1B-light chain 3II), CHOP (C/EBP homologous protein), and phospho-eIF2α (eukaryotic initiation factor 2α). Finally, bacterial artificial chromosome ATXN2-Q22 mice treated with 17-DMAG or HSP990 exhibited highly reduced ATXN2 protein abundance in the cerebellum. Taken together, our study demonstrates inhibition of HSP90 or Na+/K+-ATPases as potentially effective therapeutic strategies for treating SCA2 and ALS.

Human GPR17 missense variants identified in metabolic disease patients have distinct downstream signaling profiles.

GPR17 is a G-protein-coupled receptor (GPCR) implicated in the regulation of glucose metabolism and energy homeostasis. Such evidence is primarily drawn from mouse knockout studies and suggests GPR17 as a potential novel therapeutic target for the treatment of metabolic diseases. However, links between human GPR17 genetic variants, downstream cellular signaling, and metabolic diseases have yet to be reported. Here, we analyzed GPR17 coding sequences from control and disease cohorts consisting of individuals with adverse clinical metabolic deficits including severe insulin resistance, hypercholesterolemia, and obesity. We identified 18 nonsynonymous GPR17 variants, including eight variants that were exclusive to the disease cohort. We characterized the protein expression levels, membrane localization, and downstream signaling profiles of nine GPR17 variants (F43L, V96M, V103M, D105N, A131T, G136S, R248Q, R301H, and G354V). These nine GPR17 variants had similar protein expression and subcellular localization as wild-type GPR17; however, they showed diverse downstream signaling profiles. GPR17-G136S lost the capacity for agonist-mediated cAMP, Ca2+, and ß-arrestin signaling. GPR17-V96M retained cAMP inhibition similar to GPR17-WT, but showed impaired Ca2+ and ß-arrestin signaling. GPR17-D105N displayed impaired cAMP and Ca2+ signaling, but unaffected agonist-stimulated ß-arrestin recruitment. The identification and functional profiling of naturally occurring human GPR17 variants from individuals with metabolic diseases revealed receptor variants with diverse signaling profiles, including differential signaling perturbations that resulted in GPCR signaling bias. Our findings provide a framework for structure-function relationship studies of GPR17 signaling and metabolic disease.

The AKT modulator A-443654 reduces α-synuclein expression and normalizes ER stress and autophagy.

Accumulation of α-synuclein is a main underlying pathological feature of Parkinson's disease and α-synucleinopathies, for which lowering expression of the α-synuclein gene (SNCA) is a potential therapeutic avenue. Using a cell-based luciferase reporter of SNCA expression we performed a quantitative high-throughput screen of 155,885 compounds and identified A-443654, an inhibitor of the multiple functional kinase AKT, as a potent inhibitor of SNCA. HEK-293 cells with CAG repeat expanded ATXN2 (ATXN2-Q58 cells) have increased levels of α-synuclein. We found that A-443654 normalized levels of both SNCA mRNA and α-synuclein monomers and oligomers in ATXN2-Q58 cells. A-443654 also normalized levels of α-synuclein in fibroblasts and iPSC-derived dopaminergic neurons from a patient carrying a triplication of the SNCA gene. Analysis of autophagy and endoplasmic reticulum stress markers showed that A-443654 successfully prevented α-synuclein toxicity and restored cell function in ATXN2-Q58 cells, normalizing the levels of mTOR, LC3-II, p62, STAU1, BiP, and CHOP. A-443654 also decreased the expression of DCLK1, an inhibitor of α-synuclein lysosomal degradation. Our study identifies A-443654 and AKT inhibition as a potential strategy for reducing SNCA expression and treating Parkinson's disease pathology.

Retro Drug Design: From Target Properties to Molecular Structures.

To deliver more therapeutics to more patients more quickly and economically is the ultimate goal of pharmaceutical researchers. The advent and rapid development of artificial intelligence (AI), in combination with other powerful computational methods in drug discovery, makes this goal more practical than ever before. Here, we describe a new strategy, retro drug design, or RDD, to create novel small-molecule drugs from scratch to meet multiple predefined requirements, including biological activity against a drug target and optimal range of physicochemical and ADMET properties. The molecular structure was represented by an atom typing based molecular descriptor system, optATP, which was further transformed to the space of loading vectors from principal component analysis. Traditional predictive models were trained over experimental data for the target properties using optATP and shallow machine learning methods. The Monte Carlo sampling algorithm was then utilized to find the solutions in the space of loading vectors that have the target properties. Finally, a deep learning model was employed to decode molecular structures from the solutions. To test the feasibility of the algorithm, we challenged RDD to generate novel kinase inhibitors from random numbers with five different ADMET properties optimized at the same time. The best Tanimoto similarity score between the generated valid structures and the available 4,314 kinase inhibitors was < 0.50, indicating a high extent of novelty of the generated compounds. From the 3,040 structures that met all six target properties, 20 were selected for synthesis and experimental measurement of inhibition activity over 97 representative kinases and the ADMET properties. Fifteen and eight compounds were determined to be hits or strong hits, respectively. Five of the six strong kinase inhibitors have excellent experimental ADMET properties. The results presented in this paper illustrate that RDD has the potential to significantly improve the current drug discovery process.

Discovery of Small-Molecule VapC1 Nuclease Inhibitors by Virtual Screening and Scaffold Hopping from an Atomic Structure Revealing Protein-Protein Interactions with a Native VapB1 Inhibitor.

Nontypeable Haemophilus influenzae (NTHi) are clinically important Gram-negative bacteria that are responsible for various human mucosal diseases, including otitis media (OM). Recurrent OM caused by NTHi is common, and infections that recur less than 2 weeks following antimicrobial therapy are largely attributable to the recurrence of the same strain of bacteria. Toxin-antitoxin (TA) modules encoded by bacteria enable rapid responses to environmental stresses and are thought to facilitate growth arrest, persistence, and tolerance to antibiotics. The vapBC-1 locus of NTHi encodes a type II TA system, comprising the ribonuclease toxin VapC1 and its cognate antitoxin VapB1. The activity of VapC1 has been linked to the survival of NTHi during antibiotic treatment both in vivo and ex vivo. Therefore, inhibitors of VapC1 might serve as adjuvants to antibiotics, preventing NTHi from entering growth arrest and surviving; however, none have been reported to date. A truncated VapB1 peptide from a crystal structure of the VapBC-1 complex was used to generate pharmacophore queries to facilitate a scaffold hopping approach for the identification of small-molecule VapC1 inhibitors. The National Center for Advancing Translational Sciences small-molecule library was virtually screened using the shape-based method rapid overlay of chemical structures (ROCS), and the top-ranking hits were docked into the VapB1 binding pocket of VapC1. Two hundred virtual screening hits with the best docking scores were selected and tested in a biochemical VapC1 activity assay, which confirmed eight compounds as VapC1 inhibitors. An additional 60 compounds were selected with structural similarities to the confirmed VapC1 inhibitors, of which 20 inhibited VapC1 activity. Intracellular target engagement of five inhibitors was indicated by the destabilization of VapC1 within bacterial cells from a cellular thermal shift assay; however, no impact on bacterial growth was observed. Thus, this virtual screening and scaffold hopping approach enabled the discovery of VapC1 ribonuclease inhibitors that might serve as starting points for preclinical development.

Virtual Screening for the Discovery of Microbiome ß-Glucuronidase Inhibitors to Alleviate Cancer Drug Toxicity.

Despite the potency of most first-line anti-cancer drugs, nonadherence to these drug regimens remains high and is attributable to the prevalence of "off-target" drug effects that result in serious adverse events (SAEs) like hair loss, nausea, vomiting, and diarrhea. Some anti-cancer drugs are converted by liver uridine 5'-diphospho-glucuronosyltransferases through homeostatic host metabolism to form drug-glucuronide conjugates. These sugar-conjugated metabolites are generally inactive and can be safely excreted via the biliary system into the gastrointestinal tract. However, ß-glucuronidase (ßGUS) enzymes expressed by commensal gut bacteria can remove the glucuronic acid moiety, producing the reactivated drug and triggering dose-limiting side effects. Small-molecule ßGUS inhibitors may reduce this drug-induced gut toxicity, allowing patients to complete their full course of treatment. Herein, we report the discovery of novel chemical series of ßGUS inhibitors by structure-based virtual high-throughput screening (vHTS). We developed homology models for ßGUS and applied them to large-scale vHTS against nearly 400,000 compounds within the chemical libraries of the National Center for Advancing Translational Sciences at the National Institutes of Health. From the vHTS results, we cherry-picked 291 compounds via a multifactor prioritization procedure, providing 69 diverse compounds that exhibited positive inhibitory activity in a follow-up ßGUS biochemical assay in vitro. Our findings correspond to a hit rate of 24% and could inform the successful downstream development of a therapeutic adjunct that targets the human microbiome to prevent SAEs associated with first-line, standard-of-care anti-cancer drugs.

Deep Graph Learning with Property Augmentation for Predicting Drug-Induced Liver Injury.

Drug-induced liver injury (DILI) is a crucial factor in determining the qualification of potential drugs. However, the DILI property is excessively difficult to obtain due to the complex testing process. Consequently, an in silico screening in the early stage of drug discovery would help to reduce the total development cost by filtering those drug candidates with a high risk to cause DILI. To serve the screening goal, we apply several computational techniques to predict the DILI property, including traditional machine learning methods and graph-based deep learning techniques. While deep learning models require large training data to tune huge model parameters, the DILI data set only contains a few hundred annotated molecules. To alleviate the data scarcity problem, we propose a property augmentation strategy to include massive training data with other property information. Extensive experiments demonstrate that our proposed method significantly outperforms all existing baselines on the DILI data set by obtaining a 81.4% accuracy using cross-validation with random splitting, 78.7% using leave-one-out cross-validation, and 76.5% using cross-validation with scaffold splitting.

Identification of SARS-CoV-2 viral entry inhibitors using machine learning and cell-based pseudotyped particle assay.

In response to the pandemic caused by SARS-CoV-2, we constructed a hybrid support vector machine (SVM) classification model using a set of publicly posted SARS-CoV-2 pseudotyped particle (PP) entry assay repurposing screen data to identify novel potent compounds as a starting point for drug development to treat COVID-19 patients. Two different molecular descriptor systems, atom typing descriptors and 3D fingerprints (FPs), were employed to construct the SVM classification models. Both models achieved reasonable performance, with the area under the curve of receiver operating characteristic (AUC-ROC) of 0.84 and 0.82, respectively. The consensus prediction outperformed the two individual models with significantly improved AUC-ROC of 0.91, where the compounds with inconsistent classifications were excluded. The consensus model was then used to screen the 173,898 compounds in the NCATS annotated and diverse chemical libraries. Of the 255 compounds selected for experimental confirmation, 116 compounds exhibited inhibitory activities in the SARS-CoV-2 PP entry assay with IC50 values ranged between 0.17 µM and 62.2 µM, representing an enrichment factor of 3.2. These 116 active compounds with diverse and novel structures could potentially serve as starting points for chemistry optimization for COVID-19 drug discovery.

Physiologically relevant orthogonal assays for the discovery of small-molecule modulators of WIP1 phosphatase in high-throughput screens.

WT P53-Induced Phosphatase 1 (WIP1) is a member of the magnesium-dependent serine/threonine protein phosphatase (PPM) family and is induced by P53 in response to DNA damage. In several human cancers, the WIP1 protein is overexpressed, which is generally associated with a worse prognosis. Although WIP1 is an attractive therapeutic target, no potent, selective, and bioactive small-molecule modulator with favorable pharmacokinetics has been reported. Phosphatase enzymes are among the most challenging targets for small molecules because of the difficulty of achieving both modulator selectivity and bioavailability. Another major obstacle has been the availability of robust and physiologically relevant phosphatase assays that are suitable for high-throughput screening. Here, we describe orthogonal biochemical WIP1 activity assays that utilize phosphopeptides from native WIP1 substrates. We optimized an MS assay to quantify the enzymatically dephosphorylated peptide reaction product in a 384-well format. Additionally, a red-shifted fluorescence assay was optimized in a 1,536-well format to enable real-time WIP1 activity measurements through the detection of the orthogonal reaction product, Pi We validated these two optimized assays by quantitative high-throughput screening against the National Center for Advancing Translational Sciences (NCATS) Pharmaceutical Collection and used secondary assays to confirm and evaluate inhibitors identified in the primary screen. Five inhibitors were further tested with an orthogonal WIP1 activity assay and surface plasmon resonance binding studies. Our results validate the application of miniaturized physiologically relevant and orthogonal WIP1 activity assays to discover small-molecule modulators from high-throughput screens.

Predictive models for estimating cytotoxicity on the basis of chemical structures.

Cytotoxicity is a critical property in determining the fate of a small molecule in the drug discovery pipeline. Cytotoxic compounds are identified and triaged in both target-based and cell-based phenotypic approaches due to their off-target toxicity or on-target and on-mechanism toxicity for oncology and neurodegenerative targets. It is critical that chemical-induced cytotoxicity be reliably predicted before drug candidates advance to the late stage of development, or more ideally, before compounds are synthesized. In this study, we assessed the cell-based cytotoxicity of nearly 10,000 compounds in NCATS annotated libraries against four 'normal' cell lines (HEK 293, NIH 3T3, CRL-7250 and HaCat) using CellTiter-Glo (CTG) technology and constructed highly predictive models to estimate cytotoxicity from chemical structures. There are 5,241 non-redundant compounds having unambiguous activities in the four different cell lines, among which 11.8% compounds exhibited cytotoxicity in two or more cell lines and are thus labelled cytotoxic. The support vector classification (SVC) models trained with 80% randomly selected molecules achieved the area under the receiver operating characteristic curve (AUC-ROC) of 0.88 on average for the remaining 20% compounds in the test sets in 10 repeating experiments. Application of under-sampling rebalancing method further improved the averaged AUC-ROC to 0.90. Analysis of structural features shared by cytotoxic compounds may offer medicinal chemists heuristic design ideas to eliminate undesirable cytotoxicity. The profiling of cytotoxicity of drug-like molecules with annotated primary mechanism of action (MOA) will inform on the roles played by different targets or pathways in cellular viability. The predictive models for cytotoxicity (accessible at https://tripod.nih.gov/web_adme/cytotox.html) provide the scientific community a fast yet reliable way to prioritize molecules with little or no cytotoxicity for downstream development.

Predictive models of aqueous solubility of organic compounds built on A large dataset of high integrity.

Aqueous solubility is one of the most important properties in drug discovery, as it has profound impact on various drug properties, including biological activity, pharmacokinetics (PK), toxicity, and in vivo efficacy. Both kinetic and thermodynamic solubilities are determined during different stages of drug discovery and development. Since kinetic solubility is more relevant in preclinical drug discovery research, especially during the structure optimization process, we have developed predictive models for kinetic solubility with in-house data generated from 11,780 compounds collected from over 200 NCATS intramural research projects. This represents one of the largest kinetic solubility datasets of high quality and integrity. Based on the customized atom type descriptors, the support vector classification (SVC) models were trained on 80% of the whole dataset, and exhibited high predictive performance for estimating the solubility of the remaining 20% compounds within the test set. The values of the area under the receiver operating characteristic curve (AUC-ROC) for the compounds in the test sets reached 0.93 and 0.91, when the threshold for insoluble compounds was set to 10 and 50 µg/mL respectively. The predictive models of aqueous solubility can be used to identify insoluble compounds in drug discovery pipeline, provide design ideas for improving solubility by analyzing the atom types associated with poor solubility and prioritize compound libraries to be purchased or synthesized.

Highly predictive and interpretable models for PAMPA permeability.

Cell membrane permeability is an important determinant for oral absorption and bioavailability of a drug molecule. An in silico model predicting drug permeability is described, which is built based on a large permeability dataset of 7488 compound entries or 5435 structurally unique molecules measured by the same lab using parallel artificial membrane permeability assay (PAMPA). On the basis of customized molecular descriptors, the support vector regression (SVR) model trained with 4071 compounds with quantitative data is able to predict the remaining 1364 compounds with the qualitative data with an area under the curve of receiver operating characteristic (AUC-ROC) of 0.90. The support vector classification (SVC) model trained with half of the whole dataset comprised of both the quantitative and the qualitative data produced accurate predictions to the remaining data with the AUC-ROC of 0.88. The results suggest that the developed SVR model is highly predictive and provides medicinal chemists a useful in silico tool to facilitate design and synthesis of novel compounds with optimal drug-like properties, and thus accelerate the lead optimization in drug discovery.

A selective USP1-UAF1 inhibitor links deubiquitination to DNA damage responses.

Protein ubiquitination and deubiquitination are central to the control of a large number of cellular pathways and signaling networks in eukaryotes. Although the essential roles of ubiquitination have been established in the eukaryotic DNA damage response, the deubiquitination process remains poorly defined. Chemical probes that perturb the activity of deubiquitinases (DUBs) are needed to characterize the cellular function of deubiquitination. Here we report ML323 (2), a highly potent inhibitor of the USP1-UAF1 deubiquitinase complex with excellent selectivity against human DUBs, deSUMOylase, deneddylase and unrelated proteases. Using ML323, we interrogated deubiquitination in the cellular response to UV- and cisplatin-induced DNA damage and revealed new insights into the requirement of deubiquitination in the DNA translesion synthesis and Fanconi anemia pathways. Moreover, ML323 potentiates cisplatin cytotoxicity in non-small cell lung cancer and osteosarcoma cells. Our findings point to USP1-UAF1 as a key regulator of the DNA damage response and a target for overcoming resistance to the platinum-based anticancer drugs.

Genomic and functional characterizations of phosphodiesterase subtype 4D in human cancers.

Discovery of cancer genes through interrogation of genomic dosage is one of the major approaches in cancer research. In this study, we report that phosphodiesterase subtype 4D (PDE4D) gene was homozygously deleted in 198 cases of 5,569 primary solid tumors (3.56%), with most being internal microdeletions. Unexpectedly, the microdeletions did not result in loss of their gene products. Screening PDE4D expression in 11 different types of primary tumor samples (n = 165) with immunohistochemistry staining revealed that its protein levels were up-regulated compared with corresponding nontransformed tissues. Importantly, depletion of endogenous PDE4D with three independent shRNAs caused apoptosis and growth inhibition in multiple types of cancer cells, including breast, lung, ovary, endometrium, gastric, and melanoma, which could be rescued by reexpression of PDE4D. We further showed that antitumor events triggered by PDE4D suppression were lineage-dependently associated with Bcl-2 interacting mediator of cell death (BIM) induction and microphthalmia-associated transcription factor (MITF) down-regulation. Furthermore, ectopic expression of the PDE4D short isoform, PDE4D2, enhanced the proliferation of cancer cells both in vitro and in vivo. Moreover, treatment of cancer cells with a unique specific PDE4D inhibitor, 26B, triggered massive cell death and growth retardation. Notably, these antineoplastic effects induced by either shRNAs or small molecule occurred preferentially in cancer cells but not in nonmalignant epithelial cells. These results suggest that although targeted by genomic homozygous microdeletions, PDE4D functions as a tumor-promoting factor and represents a unique targetable enzyme of cancer cells.

Are hERG channel blockers also phospholipidosis inducers?

Both pharmacophore models of the human ether-à-go-go-related gene (hERG) channel blockers and phospholipidosis (PLD) inducers contain a hydrophobic moiety and a hydrophilic motif/positively charged center, so it is interesting to investigate the overlap between the ligand chemical spaces of both targets. We have assayed over 4000 non-redundant drug-like compounds for both their hERG inhibitory activity and PLD inducing potential in a quantitative high throughput screening (qHTS) format. Seventy-seven percent of PLD inducing compounds identified from the screening were also found to be hERG channel blockers, and 96.9% of the dually active compounds were positively charged. Among the 48 compounds that induced PLD without inhibiting hERG channel, 24 compounds (50.0%) carried steroidal structures. According to our results, hERG channel blockers and PLD inducers share a large chemical space. In addition, a positively charged hERG channel blocker will most likely induce PLD, while a steroid PLD inducer is less likely a hERG channel blocker.

Identification of potent Yes1 kinase inhibitors using a library screening approach.

Yes1 kinase has been implicated as a potential therapeutic target in a number of cancers including melanomas, breast cancers, and rhabdomyosarcomas. Described here is the development of a robust and miniaturized biochemical assay for Yes1 kinase that was applied in a high throughput screen (HTS) of kinase-focused small molecule libraries. The HTS provided 144 (17% hit rate) small molecule compounds with IC50 values in the sub-micromolar range. Three of the most potent Yes1 inhibitors were then examined in a cell-based assay for inhibition of cell survival in rhabdomyosarcoma cell lines. Homology models of Yes1 were generated in active and inactive conformations, and docking of inhibitors supports binding to the active conformation (DFG-in) of Yes1. This is the first report of a large high throughput enzymatic activity screen for identification of Yes1 kinase inhibitors, thereby elucidating the polypharmacology of a variety of small molecules and clinical candidates.

Pyrido[2,3-d]pyrimidines: discovery and preliminary SAR of a novel series of DYRK1B and DYRK1A inhibitors.

DYRK1B is a kinase over-expressed in certain cancer cells (including colon, ovarian, pancreatic, etc.). Recent publications have demonstrated inhibition of DYRK1B could be an attractive target for cancer therapy. From a data-mining effort, the team has discovered analogues of pyrido[2,3-d]pyrimidines as potent enantio-selective inhibitors of DYRK1B. Cells treated with a tool compound from this series showed the same cellular effects as down regulation of DYRK1B with siRNA. Such effects are consistent with the proposed mechanism of action. Progress of the SAR study is presented.

An electrophilic fragment screening for the development of small molecules targeting caspase-2.

Recent Alzheimer's research has shown increasing interest in the caspase-2 (Casp2) enzyme. However, the available Casp2 inhibitors, which have been pentapeptides or peptidomimetics, face challenges for use as CNS drugs. In this study, we successfully screened a 1920-compound chloroacetamide-based, electrophilic fragment library from Enamine. Our two-point dose screen identified 64 Casp2 hits, which were further evaluated in a ten-point dose-response study to assess selectivity over Casp3. We discovered compounds with inhibition values in the single-digit micromolar and sub-micromolar range, as well as up to 32-fold selectivity for Casp2 over Casp3. Target engagement analysis confirmed the covalent-irreversible binding of the selected fragments to Cys320 at the active site of Casp2. Overall, our findings lay a strong foundation for the future development of small-molecule Casp2 inhibitors.

Structure based model for the prediction of phospholipidosis induction potential of small molecules.

Drug-induced phospholipidosis (PLD), characterized by an intracellular accumulation of phospholipids and formation of concentric lamellar bodies, has raised concerns in the drug discovery community, due to its potential adverse effects. To evaluate the PLD induction potential, 4,161 nonredundant drug-like molecules from the National Institutes of Health Chemical Genomics Center (NCGC) Pharmaceutical Collection (NPC), the Library of Pharmacologically Active Compounds (LOPAC), and the Tocris Biosciences collection were screened in a quantitative high-throughput screening (qHTS) format. The potential of drug-lipid complex formation can be linked directly to the structures of drug molecules, and many PLD inducing drugs were found to share common structural features. Support vector machine (SVM) models were constructed by using customized atom types or Molecular Operating Environment (MOE) 2D descriptors as structural descriptors. Either the compounds from LOPAC or randomly selected from the entire data set were used as the training set. The impact of training data with biased structural features and the impact of molecule descriptors emphasizing whole-molecule properties or detailed functional groups at the atom level on model performance were analyzed and discussed. Rebalancing strategies were applied to improve the predictive power of the SVM models. Using the undersampling method, the consensus model using one-third of the compounds randomly selected from the data set as the training set achieved high accuracy of 0.90 in predicting the remaining two-thirds of the compounds constituting the test set, as measured by the area under the receiver operator characteristic curve (AUC-ROC).

Exploring polypharmacology using a ROCS-based target fishing approach.

Polypharmacology has emerged as a new theme in drug discovery. In this paper, we studied polypharmacology using a ligand-based target fishing (LBTF) protocol. To implement the protocol, we first generated a chemogenomic database that links individual protein targets with a specified set of drugs or target representatives. Target profiles were then generated for a given query molecule by computing maximal shape/chemistry overlap between the query molecule and the drug sets assigned to each protein target. The overlap was computed using the program ROCS (Rapid Overlay of Chemical Structures). We validated this approach using the Directory of Useful Decoys (DUD). DUD contains 2950 active compounds, each with 36 property-matched decoys, against 40 protein targets. We chose a set of known drugs to represent each DUD target, and we carried out ligand-based virtual screens using data sets of DUD actives seeded into DUD decoys for each target. We computed Receiver Operator Characteristic (ROC) curves and associated area under the curve (AUC) values. For the majority of targets studied, the AUC values were significantly better than for the case of a random selection of compounds. In a second test, the method successfully identified off-targets for drugs such as rimantadine, propranolol, and domperidone that were consistent with those identified by recent experiments. The results from our ROCS-based target fishing approach are promising and have potential application in drug repurposing for single and multiple targets, identifying targets for orphan compounds, and adverse effect prediction.