Discovery of pyrazolo[1,5-a]pyrimidine-3-carbonitrile derivatives as a new class of histone lysine demethylase 4D (KDM4D) inhibitors.

Herein we report the discovery of a series of new small molecule inhibitors of histone lysine demethylase 4D (KDM4D). Molecular docking was first performed to screen for new KDM4D inhibitors from various chemical databases. Two hit compounds were retrieved. Further structural optimization and structure-activity relationship (SAR) analysis were carried out to the more selective one, compound 2, which led to the discovery of several new KDM4D inhibitors. Among them, compound 10r is the most potent one with an IC50 value of 0.41±0.03µM against KDM4D. Overall, compound 10r could be taken as a good lead compound for further studies.

A preclinical evaluation of SKLB261, a multikinase inhibitor of EGFR/Src/VEGFR2, as a therapeutic agent against pancreatic cancer.

The clinical prognosis of pancreatic cancer remains rather disappointing despite tremendous efforts in exploring medical treatments in the past two decades. Development of more effective treatment strategies is still desperately needed to improve outcomes in patients with pancreatic cancer. SKLB261 is a multikinase inhibitor obtained recently through a lead optimization. In this investigation, we shall evaluate its anti-pancreatic cancer effects both in vitro and in vivo. SKLB261 is a multikinase inhibitor potently inhibiting EGFR, Src, and VEGFR2 kinases. It could significantly inhibit cell proliferation, migration, and invasion, and induce apoptosis in cellular assays of human pancreatic cancer cells that are sensitive or resistant to dasatinib and/or gemcitabine. Western blot analysis showed that SKLB261 inhibited the activation of EGFR and Src kinases as well as their downstream signaling proteins, including FAK, ERK, and STAT3. SKLB261 also showed potent antiangiogenic effects in transgenic zebrafish models. In vivo, SKLB261 displayed more potent antitumor activities than dasatinib, gemcitabine, or erlotinib in pancreatic cancer xenografts, including BxPC-3, PANC-1, AsPC-1, and HPAC. Furthermore, mice receiving SKLB261 therapy showed significant survival advantage compared with vehicle-treated and gemcitabine-treated groups in an experimental metastasis model of pancreatic cancer. These data, together with the good pharmacokinetic properties and low toxicity of this compound, provide a rationale for the ongoing clinical evaluation of SKLB261 in the treatment of pancreatic cancer.

Virtual screening in small molecule discovery for epigenetic targets.

Epigenetic modifications are critical mechanisms that regulate many biological processes and establish normal cellular phenotypes. Aberrant epigenetic modifications are frequently linked to the development and maintenance of several diseases including cancer, inflammation and metabolic diseases and so on. The key proteins that mediate epigenetic modifications have been thus recognized as potential therapeutic targets for these diseases. Consequently, discovery of small molecule inhibitors for epigenetic targets has received considerable attention in recent years. Here, virtual screening methods and their applications in the discovery of epigenetic target inhibitors are the focus of this review. Newly emerging approaches or strategies including rescoring methods, docking pose filtering methods, machine learning methods and 3D molecular similarity methods were also underlined. They are expected to be employed for identifying novel inhibitors targeting epigenetic regulation more efficiently.

TS-Chemscore, a Target-Specific Scoring Function, Significantly Improves the Performance of Scoring in Virtual Screening.

Most of the scoring functions currently used in structure-based drug design belong to 'universal' scoring functions, which often give a poor correlation between the calculated scores and experimental binding affinities. In this investigation, we proposed a simple strategy to construct target-specific scoring functions based on known 'universal' scoring functions. This strategy was applied to Chemscore, a widely used empirical scoring function, which led to a new scoring function, termed TS-Chemscore. TS-Chemscore was validated on 14 protein targets, which cover a wide range of biological target categories. The results showed that TS-Chemscore significantly improved the correlation between the calculated scores and experimental binding affinities compared with the original Chemscore. TS-Chemscore was then applied in virtual screening to retrieve novel JAK3 and YopH inhibitors. Top 30 compounds for each target were selected for experimental validation. Six active compounds for JAK3 and four for YopH were obtained. These compounds were out of the lists of top 30 compounds sorted by Chemscore. Collectively, TS-Chemscore established in this study showed a better performance in virtual screening than its counterpart Chemscore.

A preclinical evaluation of a novel multikinase inhibitor, SKLB-329, as a therapeutic agent against hepatocellular carcinoma.

Hepatocellular carcinoma (HCC) is a serious life-threatening malignant disease of liver. Molecular targeted therapies are considered a promising strategy for the treatment of HCC. Sorafenib is the first, and so far the only targeted drug approved by the US Food and Drug Administration (FDA) for clinical therapy of HCC. Despite being effective in some HCC patients, some demerits of sorafenib in the treatment of HCC, such as modest survival benefits, and drug resistance, have also been reported, which highlights the unmet medical need among patients with HCC. Here, we report a novel multikinase inhibitor discovered by us, SKLB-329, which potently inhibits angiogenesis-related kinases including VEGFR1/2/3, and FGFR2, and the Src kinase. SKLB-329 significantly inhibited endothelial cell growth, migration, invasion and tube formation. It showed potent anti-angiogenic activity in a transgenic zebrafish model. Moreover, SKLB-329 could efficiently restrain the proliferation of HCC cells through down-regulation of Src-mediated FAK and Stat3 activity. In vivo, oral administration of SKLB-329 considerably suppressed the tumor growth in HCC xenograft models (HepG2 and SMMC7721) in a dose-dependent manner. In all of the in vitro and in vivo assays of this investigation, sorafenib was used as a positive control, and in most assays SKLB-329 exhibited a higher potency compared with the positive control. In addition, SKLB-329 also bears favorable pharmacokinetic properties. Collectively, the results of preclinical studies presented here demonstrate that SKLB-329 is a promising drug candidate for HCC treatment.

Discovery and structure-activity relationships study of novel thieno[2,3-b]pyridine analogues as hepatitis C virus inhibitors.

Current treatment for hepatitis C is barely satisfactory, there is an urgent need to develop novel agents for combating hepatitis C virus infection. This study discovered a new class of thieno[2,3-b]pyridine derivatives as HCV inhibitors. First, a hit compound characterized by a thienopyridine core was identified in a cell-based screening of our privileged small molecule library. And then, structure activity relationship study of the hit compound led to the discovery of several potent compounds without obvious cytotoxicity in vitro (12c, EC50=3.3µM, SI >30.3, 12b, EC50=3.5µM, SI >28.6, 10l, EC50=3.9µM, SI >25.6, 12o, EC50=4.5µM, SI >22.2, respectively). Although the mechanism of them had not been clearly elucidated, our preliminary optimization of this class of compounds had provided us a start point to develop new anti-HCV agents.

Retrieving novel C5aR antagonists using a hybrid ligand-based virtual screening protocol based on SVM classification and pharmacophore models.

C5aR antagonists have been thought as potential immune mediators in various inflammatory and autoimmune diseases, and discovery of C5aR antagonists has attracted much attention in recent years. The discovery of C5aR antagonists was usually achieved through high-throughput screening, which usually suffered a high cost and a low success rate. Currently, the fast developing computer-aided virtual screening (VS) methods provide economic and rapid approaches to the lead discovery. In this account, we proposed a hybrid ligand-based VS protocol that is based on support vector machine (SVM) classification and pharmacophore models for retrieving novel C5aR antagonists. Performance evaluation of this hybrid VS protocol in virtual screening against a large independent test set, T-CHEM, showed that the hybrid VS approach significantly increased the hit rate and enrichment factor compared with the individual SVM classification model-based VS and pharmacophore model-based VS, as well as molecular docking-based VS in that the receptor structure was created by homology modeling. The hybrid VS approach was then used to screen several large chemical libraries including PubChem, Specs, and Enamine. Finally, a total of 20 compounds were selected from the top ranking hits, and shifted to the subsequent in vitro and in vivo studies, which results will be reported in the near future.

Receptor-based pharmacophore and pharmacophore key descriptors for virtual screening and QSAR modeling.

The intuitive nature of the pharmacophore concept has made it widely accepted by the medicinal chemistry community, evidenced by the past 3 decades of development and application of computerized pharmacophore modeling tools. On the other hand, shape complementarity has been recognized as a critical factor in molecular recognition between drugs and their receptors. Recent development of fast and accurate shape comparison tools has facilitated the wide spread use of shape matching technologies in drug discovery. However, pharmacophore and shape technologies, if used separately, often lead to high false positive rate. Thus, integrating pharmacophore matching and shape matching technologies into one program has the potential to reduce the false positive rates in virtual screening. Other issues of current pharmacophore technologies include sometimes high false negative rate and non-quantitative prediction. In this article, we first focus on a recently implemented method (Shape4) that combines receptor based shape matching and pharmacophore comparison in a single algorithm to create shape pharmacophore models for virtual screening. We also examine a recent example that utilizes multi-complex information to develop receptor-based pharmacophore models that promises to reduce false negative rate. Finally, we review several methods that employ receptor-based pharmacophore map and pharmacophore key descriptors for QSAR modeling. We conclude by emphasizing the concept of receptor-based shape pharmacophore and its roles in future drug discovery.

Rapid and accurate assessment of seizure liability of drugs by using an optimal support vector machine method.

Drug-induced seizures are a serious adverse effect and assessment of seizure risk usually takes place at the late stage of drug discovery process, which does not allow sufficient time to reduce the risk by chemical modification. Thus early identification of chemicals with seizure liability using rapid and cheaper approaches would be preferable. In this study, an optimal support vector machine (SVM) modeling method has been employed to develop a prediction model of seizure liability of chemicals. A set of 680 compounds were used to train the SVM model. The established SVM model was then validated by an independent test set comprising 175 compounds, which gave a prediction accuracy of 86.9%. Further, the SVM-based prediction model of seizure liability was compared with various preclinical seizure assays, including in vitro rat hippocampal brain slice, in vivo zebrafish larvae assay, mouse spontaneous seizure model, and mouse EEG model. In terms of predictability, the SVM model was ranked just behind the mouse EEG model, but better than the rat brain slice and zebrafish models. Nevertheless, the SVM model has considerable advantages compared with the preclinical seizure assays in speed and cost. In summary, the SVM-based prediction model of seizure liability established here offers potential as a cheaper, rapid and accurate assessment of seizure liability of drugs, which could be used in the seizure risk assessment at the early stage of drug discovery. The prediction model is freely available online at http://www.sklb.scu.edu.cn/lab/yangsy/download/ADMET/seizure_pred.tar.

Pharmacophore modeling and hybrid virtual screening for the discovery of novel IκB kinase 2 (IKK2) inhibitors.

IKK2 (IκB kinase 2) inhibitors have been identified as potential drug candidates in the treatment of various immune/inflammatory disorders as well as cancer. So far more than one hundred small molecule inhibitors against IKK2 have been reported publicly. In this investigation, pharmacophore modeling was carried out to clarify the essential structure-activity relationship for the known IKK2 inhibitors. One of the established pharmacophore hypotheses, namely Hypo8, which has the best prediction ability to an external test data set, was suggested as a template for virtual screening. Evaluation of the performances of Hypo8 and a hybrid method (Hypo81docking) in virtual screening indicated that the use of the hybrid virtual screening considerably increased the hit rate and enrichment factor. The hybrid method was therefore adopted for screening several commercially available chemical databases, including Specs, NCI, Maybridge and Chinese Nature Product Database (CNPD), for novel potent IKK2 inhibitors. The hit compounds were subsequently subjected to filtering by Lipinski's rule of five. Finally some of the final hit compounds were selected and suggested for further experimental investigations.

Discovery of novel Pim-1 kinase inhibitors by a hierarchical multistage virtual screening approach based on SVM model, pharmacophore, and molecular docking.

In this investigation, we describe the discovery of novel potent Pim-1 inhibitors by employing a proposed hierarchical multistage virtual screening (VS) approach, which is based on support vector machine-based (SVM-based VS or SB-VS), pharmacophore-based VS (PB-VS), and docking-based VS (DB-VS) methods. In this approach, the three VS methods are applied in an increasing order of complexity so that the first filter (SB-VS) is fast and simple, while successive ones (PB-VS and DB-VS) are more time-consuming but are applied only to a small subset of the entire database. Evaluation of this approach indicates that it can be used to screen a large chemical library rapidly with a high hit rate and a high enrichment factor. This approach was then applied to screen several large chemical libraries, including PubChem, Specs, and Enamine as well as an in-house database. From the final hits, 47 compounds were selected for further in vitro Pim-1 inhibitory assay, and 15 compounds show nanomolar level or low micromolar inhibition potency against Pim-1. In particular, four of them were found to have new scaffolds which have potential for the chemical development of Pim-1 inhibitors.

Effect of training data size and noise level on support vector machines virtual screening of genotoxic compounds from large compound libraries.

Various in vitro and in-silico methods have been used for drug genotoxicity tests, which show limited genotoxicity (GT+) and non-genotoxicity (GT-) identification rates. New methods and combinatorial approaches have been explored for enhanced collective identification capability. The rates of in-silco methods may be further improved by significantly diversified training data enriched by the large number of recently reported GT+ and GT- compounds, but a major concern is the increased noise levels arising from high false-positive rates of in vitro data. In this work, we evaluated the effect of training data size and noise level on the performance of support vector machines (SVM) method known to tolerate high noise levels in training data. Two SVMs of different diversity/noise levels were developed and tested. H-SVM trained by higher diversity higher noise data (GT+ in any in vivo or in vitro test) outperforms L-SVM trained by lower noise lower diversity data (GT+ in in vivo or Ames test only). H-SVM trained by 4,763 GT+ compounds reported before 2008 and 8,232 GT- compounds excluding clinical trial drugs correctly identified 81.6% of the 38 GT+ compounds reported since 2008, predicted 83.1% of the 2,008 clinical trial drugs as GT-, and 23.96% of 168 K MDDR and 27.23% of 17.86M PubChem compounds as GT+. These are comparable to the 43.1-51.9% GT+ and 75-93% GT- rates of existing in-silico methods, 58.8% GT+ and 79% GT- rates of Ames method, and the estimated percentages of 23% in vivo and 31-33% in vitro GT+ compounds in the "universe of chemicals". There is a substantial level of agreement between H-SVM and L-SVM predicted GT+ and GT- MDDR compounds and the prediction from TOPKAT. SVM showed good potential in identifying GT+ compounds from large compound libraries based on higher diversity and higher noise training data.

[The feasibility of application of reverse docking method to the selectivity studies of protein kinase inhibitors].

This investigation is to explore the feasibility of applying reverse docking method to the selectivity studies of protein kinase inhibitors. Firstly, a database that consists of 422 protein kinase structures was established through collecting the reported crystal structures or homology modeling. Then a reverse docking based method of protein kinase target screening was established, followed by the optimization of related parameters and scoring functions. Finally, seven typical selective kinase inhibitors were used to test the established method. The results show that the selective targets of these inhibitors have relatively high scoring function values (ranking in the first 35% of the tested kinase targets according to the scoring function values). This implies that the reverse docking method can be applied to the virtual screening of kinase targets and further to the selectivity studies of protein kinase inhibitors.

A specific pharmacophore model of Aurora B kinase inhibitors and virtual screening studies based on it.

In this study, 3D-pharmacophore models of Aurora B kinase inhibitors have been developed by using HipHop and HypoGen modules in Catalyst software package. The best pharmacophore model, Hypo1, which has the highest correlation coefficient (0.9911), consists of one hydrogen-bond acceptor, one hydrogen-bond donor, one hydrophobic aliphatic moiety and one ring aromatic feature. Hypo1 was validated by test set and cross-validation methods. And the specificity of Hypo1 to Aurora B inhibitors was examined with the use of selective inhibitors against Aurora B and its paralogue Aurora A. The results clearly indicate that Hypo1 can differentiate selective inhibitors of Aurora B from those of Aurora A, and the ring aromatic feature likely plays some important roles for the specificity of Hypo1. Then Hypo1 was used as a 3D query to screen several databases including Specs, NCI, Maybridge and Chinese Nature Product Database (CNPD) for identifying new inhibitors of Aurora B. The hit compounds were subsequently subjected to filtering by Lipinski's rule of five and docking studies to refine the retrieved hits, and some compounds selected from the top ranked hits have been suggested for further experimental assay studies.

In silico prediction of mitochondrial toxicity by using GA-CG-SVM approach.

Drug-induced mitochondrial toxicity has become one of the key reasons for which some drugs fail to enter market or are withdrawn from market. Thus early identification of new chemical entities that injure mitochondrial function grows to be very necessary to produce safer drugs and directly reduce attrition rate in later stages of drug development. In this study, support vector machine (SVM) method combined with genetic algorithm (GA) for feature selection and conjugate gradient method (CG) for parameter optimization (GA-CG-SVM), has been employed to develop prediction model of mitochondrial toxicity. We firstly collected 288 compounds, including 171 MT+ and 117 MT-, from different literature resources. Then these compounds were randomly separated into a training set (253 compounds) and a test set (35 compounds). The overall prediction accuracy for the training set by means of 5-fold cross-validation is 84.59%. Further, the SVM model was evaluated by using the independent test set. The overall prediction accuracy for the test set is 77.14%. These clearly indicate that the mitochondrial toxicity is predictable. Meanwhile impacts of the feature selection and SVM parameter optimization on the quality of SVM model were also examined and discussed. The results implicate the potential of the proposed GA-CG-SVM in facilitating the prediction of mitochondrial toxicity.

Pharmacophore modeling and virtual screening for designing potential PLK1 inhibitors.

Pharmacophore models of Polo-like kinase-1 (PLK1) inhibitors have been established by using the HipHop and HypoGen algorithms implemented in the Catalyst software package. The best quantitative pharmacophore model, Hypo1, which has the highest correlation coefficient (0.9895), consists of one hydrogen bond acceptor, one hydrogen bond donor, one hydrophobic feature, and one hydrophobic aliphatic feature. Hypo1 was further validated by test set and cross validation method. Then Hypo1 was used as a 3D query to screen several databases including Specs, NCI, Maybridge, and Chinese Nature Product Database (CNPD). The hit compounds were subsequently subjected to filtering by Lipinski's rule of five and docking study to refine the retrieved hits and as a result to reduce the rate of false positive. Finally, a total of 20 compounds were selected and have been shifted to in vitro and in vivo studies. As far as we know, this is the first report on the pharmacophore modeling even the first publicly reported virtual screening study of PLK1 inhibitors.

Pharmacophore modelling and virtual screening for identification of new Aurora-A kinase inhibitors.

Aurora-A has been identified as one of the most attractive targets for cancer therapy and a considerable number of Aurora-A inhibitors have been reported recently. In order to clarify the essential structure-activity relationship for the known Aurora-A inhibitors as well as identify new lead compounds against Aurora-A, 3D pharmacophore models were developed based on the known inhibitors. The best hypothesis, Hypo1, was used to screen molecular structural databases, including Specs and China Natural Products Database for potential lead compounds. The hit compounds were subsequently subjected to filtering by Lipinski's rules and docking study to refine the retrieved hits and as a result to reduce the rate of false positive. Finally, 39 compounds were purchased for further in vitro assay against several human tumour cell lines including A549, MCF-7, HepG2 and PC-3, in which Aurora-A is overexpressed. Two compounds show very low micromolar inhibition potency against some of these tumour cells. And they have been selected for further investigation.