Urumqi where can I buy cocaine

__________________________

📍 Verified store!

📍 Guarantees! Quality! Reviews!

__________________________

▼▼ ▼▼ ▼▼ ▼▼ ▼▼ ▼▼ ▼▼

▲▲ ▲▲ ▲▲ ▲▲ ▲▲ ▲▲ ▲▲

Urumqi where can I buy cocaine

Contact Us. How should we reach out? Email Me Call Me. First Name. Last Name. Email Address. Phone Number. What time works best? How did you hear about us? Your Message. Join the Boozeletter. Posted by Stoke Media Team 5 years ago March 5, Who said Pablo Escobar needed to be involved to get you high? The drug of Kings, alcohol has been creating love, causing fights and drowning sorrows since BC. Like a good suit, it never goes out of fashion. Spanish wine drinking experience? Go home, crack the kettle on, get your mates around and enjoy a game of Monopoly with a hearty cup of tea. You got done. Now we will never condone cigarettes. And I really mean at least. Pills, powder, and grass. Until they hurt, and then they really hurt. How about you just come on down to Barcelona, we can have a few shandies together and see where the morning takes us? Sign up here to receive travel dates, insider info, and travel gossip. Read More. These little…. By KP, our resident food critic Surely you know that StokeTravel do a bottomless brunch with unlimited mimosas at all of our destinations, because we know that nothing beats a…. By KP, our resident food critic Italian food is a classic favourite amongst the masses. With thriving Indian, Pakistani and Thai communities, there are plenty…. But between the gems there are lots of lemons — particularly in the more touristy centre…. Everyone knows Europe is the place to study abroad, but it can be hard to decide what country, let alone city, you should spend the next few months in. Kicking off on July…. Pride , boat party , Barcelona , party , booze , Spain. Got a question? Let's Chat On WhatsApp. Stoke Travel. Email info stoketravel. Share this post:. Join the Boozeletter Sign up here to receive travel dates, insider info, and travel gossip. Connect With Us. Related Trips. Barcelona Party Bus. View Trip. Barcelona Rooftop Craft Beer Tastings. Barcelona Wine Tours And Tastings. Barcelona Cocktail Class. The Spanish. Barcelona City Break Barcelona Boat Party and Cruises. Barcelona Cooking Class. San Juan. Related Articles. The Best Brunch In Barcelona By KP, our resident food critic Surely you know that StokeTravel do a bottomless brunch with unlimited mimosas at all of our destinations, because we know that nothing beats a…. Consenting to these technologies will allow us to process data such as browsing behavior or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions. Functional Functional Always active The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network. The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user. The technical storage or access that is used exclusively for statistical purposes. The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you. The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes. Accept Deny View preferences Save preferences View preferences. Manage Consent.

The Easy to Find Drugs in Barcelona

Urumqi where can I buy cocaine

Identifying new indications for drugs plays an essential role at many phases of drug research and development. Computational methods are regarded as an effective way to associate drugs with new indications. However, most of them complete their tasks by constructing a variety of heterogeneous networks without considering the biological knowledge of drugs and diseases, which are believed to be useful for improving the accuracy of drug repositioning. To this end, a novel heterogeneous information network HIN based model, namely HINGRL, is proposed to precisely identify new indications for drugs based on graph representation learning techniques. More specifically, HINGRL first constructs a HIN by integrating drug—disease, drug—protein and protein—disease biological networks with the biological knowledge of drugs and diseases. Then, different representation strategies are applied to learn the features of nodes in the HIN from the topological and biological perspectives. Finally, HINGRL adopts a Random Forest classifier to predict unknown drug—disease associations based on the integrated features of drugs and diseases obtained in the previous step. Experimental results demonstrate that HINGRL achieves the best performance on two real datasets when compared with state-of-the-art models. Besides, our case studies indicate that the simultaneous consideration of network topology and biological knowledge of drugs and diseases allows HINGRL to precisely predict drug—disease associations from a more comprehensive perspective. The traditional process of drug discovery suffers from the disadvantages of being labor-intensive, time-consuming and high-risk. In this regard, drug repositioning has attracted increasing attention in the pharmaceutical industry, and has achieved successful applications over the past years. For example, sildenafil was originally utilized to treat the cardiovascular disease, but later it was found to have an effect on the erectile function of male patients \[ 4 \]. Traditional drug repositioning approaches target to find abnormal clinical manifestations by manually screening clinical drug databases, and they require a large number of testing experiments on the targeted drugs. Recently, due to the increased accumulation of high-throughput genomics and proteomics data, much more attention has been given to develop different computational methods based on data mining techniques \[ 5 \]. The main reason for the popularity of computational methods is because of its advantages of low-cost and high-efficiency. At present, existing computational methods proposed for drug repositioning are classified into four categories, including recommender system-based methods, machine learning-based methods, deep learning-based methods and network-based methods \[ 6 \]. Recommender system-based methods consider the identification of potential drug indications as a recommendation task and mainly adopt the matrix factorization approach to complete their tasks \[ 7—10 \]. Although effective, these methods are not applicable to make an accurate prediction for new drugs or diseases. Machine learning-based methods are widely applied to predict associations between drugs and diseases \[ 11 , 12 \]. However, they heavily rely on the input data that is assumed to well represent the characteristics of drugs and diseases, and such assumption is difficult to satisfy in practical applications. Taking advantage of its powerful learning ability, deep learning-based methods can directly transform the original data into abstract feature representation \[ 13 , 14 \]. Although they are able to address the incompleteness problem of manually curated features \[ 15 \], a large amount of training data is required for them to obtain high accuracy. In other words, deep learning-based methods are prone to over-fitting if the input drug—disease association network is sparse. Network-based methods are widely applied for drug repositioning \[ 16—19 \]. Their performances have been verified to be better than those in the other three categories, as they improve the accuracy of drug repositioning by capturing similar information across different kinds of biological networks as the features of drugs and diseases \[ 20 \]. To do so, heterogeneous networks are introduced to represent the integration of different kinds of biological networks, and the similarities preserved across different biological networks gain new insight into the prediction of unobserved associations between drugs and diseases. However, network-based methods concentrate on constructing various heterogeneous networks while ignoring the intrinsic characteristics of different kinds of molecules, thus making it difficult to fully exploit the potential knowledge of biological networks for accurate drug repositioning. Previous studies have shown that the additional consideration of node attributes is of great significance in conducting an accuracy analysis for complex networks \[ 21—25 \], but few attempts have been made in drug repositioning by simultaneously considering network topology and biological knowledge of drugs and diseases in the same heterogeneous network. A major reason for that phenomenon is the lack of a general model that possesses the ability of properly handling these two kinds of information for predicting the associations between drugs and diseases. Furthermore, most of existing drug repositioning methods ignore the critical role of proteins when discovering novel associations between drugs and diseases. As has been pointed out by \[ 26 \], proteins are an active macromolecule in biological cells. The change in protein expressions is directly related to disease manifestation and drug action. Specifically, drugs improve disease symptoms by acting on enzymes in living organisms. Taking valproic acid as an example, the expression of histone proteins is affected in cells, thus changing the life cycle of breast cancer cells \[ 27 \]. In this regard, it is of great significance to introduce proteins to predict the relationship between drugs and diseases. Moreover, giving the fact that biological networks composed of drugs and diseases are normally sparse, the connectivity between drugs and diseases can thus be enhanced if protein—drug and protein—diseases association are integrated into these networks. To address these challenges, a novel model, namely HINGRL, is proposed to integrate network topology and biological knowledge of drugs and diseases for drug repositioning. To distinguish from existing network-based methods that focus on heterogeneous networks, a heterogeneous information network HIN is introduced for the additional consideration of biological knowledge. More specifically, HINGRL first integrates three kinds of biological networks including drug—disease, drug—protein and protein—disease networks, to obtain a HIN with the biological information of drugs and diseases collected from drug structures and semantic knowledge graphs of disease, respectively. After that, different representation learning techniques are adopted by HINGRL to learn the features of nodes in the HIN from the topological and biological perspectives. In particular, the biological knowledge of drugs and diseases is processed by using different metrics in order to obtain similarity matrices and then autoencoders are applied to construct the biological feature vectors of drugs and diseases in a more concise manner. To properly handling the information of network topology, a well-established graph representation learning algorithm, i. DeepWalk, is adopted such that the network representations of drugs and diseases can be learned from the topological perspective. After that, the biological and topological representations of drugs and diseases obtained from the given HIN are concatenated together to compose integrated feature vectors of drugs and diseases, which are then considered as the input of a Random Forest RF classifier to complete the task of predicting potential drug—disease associations. Experimental results demonstrate that HINGRL performs better in terms of several independent metrics on two real datasets when compared with state-of-the-art prediction models proposed for drug repositioning. The main contributions of this work are summarized as:. Rich heterogeneous information, i. Different graph representation learning techniques are adopted by HINGRL to better learn the integrated features of drugs and diseases by simultaneous considering network topology and biological knowledge of drugs and diseases. Experimental results demonstrate that HINGRL outperforms several state-of-the-art algorithms on two benchmark datasets of drug repositioning. To construct a HIN for performance evaluation, we adopt a benchmark dataset, namely B-dataset, composed of three kinds of biological association networks, including drug—disease, drug—protein and protein—disease associations. Among them, the drug—disease association network is obtained from the CTD database \[ 28 \] by Zhang et al. The drug—protein association network is collected from the DrugBank database \[ 30 \], and it is composed of drugs, proteins and 11 verified drug—protein associations. The protein—disease association network is derived from the DisGeNET database \[ 31 \], and there are proteins, diseases and 25 protein—disease associations in it. Moreover, to better demonstrate the generalization ability of HINGRL, we also evaluate its performance on another benchmark dataset, namely F-dataset, obtained from Gottlieb et al. F-dataset is much sparser than B-dataset in terms of the amount of drug—disease associations, as it only includes drugs, diseases and drug—disease interactions. By scanning these two databases, a total of drug—protein associations and 71 protein—disease associations are collected to compose the drug—protein and protein—disease association networks. To construct the set of negative samples from B-dataset and F-dataset, HINGRL randomly pairs up drugs and diseases whose associations are not found in the positive samples, and moreover the number of negative samples is equal to that of positive samples to avoid the unbalanced issue. As mentioned before, a HIN of interest is composed by drug—disease, drug—protein and protein—disease association networks. Obviously, there are two kinds of information available in the HIN, one is the biological knowledge of drugs and diseases and the other is the network topology. To model a HIN, we introduce a three-element tuple, i. Regarding the biological knowledge of drugs, since drug molecules with similar chemical structures are normally involved in the same biological activities \[ 32 \], we make use of such chemical information as the biological attributes of drugs. After that, we adopt the RDKit \[ 34 \] tool to examine the existence of a particular chemical structure in drug molecules. It is worth noting that in the F-dataset, since the identifiers of diseases are not consistent with those used by MeSH, we could not able to obtain their MeSH descriptors. The advantage of using autoencoder is that it solves the problem of redundancy and sparsity in the original data. In this regard, it is anticipated to not only improve the generalization ability of HINGRL but also avoid the overfitting during training. In autoencoder, there are three layers including input layer, hidden layer and output layer. Specifically, the input and output layers denote the original and new feature spaces, respectively, whereas the hidden layer is to ensure that the loss in the conversion from the original space to the new one is minimized. Moreover, the loss functions of biological knowledge extraction and network representations of drugs and diseases are presented as Equations 7 and 13 , respectively. During the training phase, pairs of drugs and diseases compose the training dataset. For each pair, the representation vectors of its drug and disease are combined as the input of RF. It is plotted by two variables including false positive rate and true positive rate. Considering the biased performance of arear under the curve AUC for imbalanced datasets, we also make use of the precision—recall PR curve to precisely reflect the actual performance of prediction models. Another two indicators, i. Matthews correlation coefficient MCC and F1-score, are also used to evaluate the overall performance of prediction models from different perspectives. Taking B-dataset as an example, we first split it into split into folds. This procedure is repeated for 10 times by alternatively taking each fold as testing data. For the purpose of performance evaluation, we compare HINGRL with three state-of-the-art algorithms proposed for drug repositioning, i. Among them, LAGCN learns the embeddings of drugs and diseases from multiple networks through a graph convolution algorithm, and then adopts attention mechanisms to integrate these embeddings for predicting new associations. DTINet obtains the characteristic representations of drugs and proteins from different biological networks, and then searches for an optimal projection to force the feature vectors of drugs close to the known interacting proteins in the space. For deepDR, multiple drug-related heterogeneous networks are constructed to extract the features of drugs during repurposing, and then utilizes the random walk with restart algorithm to infer the potential indications of drugs by capturing the representations of these networks. One should note that all these three competing algorithms make use of drug—disease associations, but LAGCN additionally integrates the biological knowledge of drugs and diseases during repurposing. Regarding the setting of parameters involved when running these algorithms, we adopt the default parameter settings for the competing models, i. Meanwhile, we conduct several trials with different settings and take the parameter values that obtain the best performance of HINGRL as the recommended setting. One should also note that all competing models are re-trained on each dataset by using the default parameter settings. This could be a strong indicator that HINGRL is preferred over state-of-the-art algorithms when applied to drug repositioning. In addition to its superior accuracy, HINGRL is also more robust than the other algorithms as indicated by their evaluation scores. In other words, HINGRL is preferred over competing models in terms of the ability of discovering novel drug—disease association as indicated by its superior performance in terms of Recall. But for HINGRL, its performance fluctuation across all the evaluation metrics is much less than the other three algorithms. First, the introduction of heterogeneous information allows HINGRL to predict unknown drug—disease associations from different perspectives. The main reason for that phenomenon is due to the imbalance in our benchmark datasets, where the number of positive samples is much less than that of negative samples. Moreover, the sparsity of HIN also accounts for the unsatisfactory performance of LAGCN in all the evaluation metrics except AUC, and accordingly the graph convolutional network used by LAGCN tends to over smooth when learning the representation from drug—disease association networks. But for HINGRL, the influence of sparsity is alleviated by using graph embedding, which is able to learn the representation of drugs and diseases from the perspective of network topology in a more effective way. The reasons for that phenomenon are two-fold: 1 the HIN constructed from F-dataset is much sparser than that from B-dataset, and accordingly fewer overlapping nodes are observed in the random walk sequences involved in F-dataset; 2 after visualizing both B-dataset and F-dataset, we find that the modularity of the HIN constructed from F-dataset is better than that from B-dataset, thus making HINGRL able to learn the topological representation of nodes in a more effective manner. Nevertheless, the introduction of heterogeneous information provides us an alternative view to complete the task of drug repositioning even some information is missed, thus enhancing the robustness of HINGRL. The RF classifiers used by these two variants are configured with the same parameters and their performances are also evaluated under fold CV. In other words, only relying on the biological knowledge of drugs and diseases may not be sufficiently enough to achieve a promising performance for drug repositioning. Since new diseases often encounter the situation that no associations are verified with existing drugs, this could be a strong indicator that HINGRL is particularly useful to identify novel indications for new drugs by only making use of their biological information. Hence, the network topology information represented by drug—disease associations allows HINGRL-B to better capture the characteristics of drugs and diseases when training the RF classifier. Lastly, a further improvement is observed from HINGRL by taking into account more heterogeneous association information, i. To this end, experiments have been conducted by comparing the performance of HINGRL with the use of different classifiers. Regarding the hyperparameters setting of each machine learning algorithm, taking the KNN classifier as an example, the number of neighbors is of great significance to tune the performance of KNN and hence we conduct several trials by varying its value from 1 to 14 at a step size of 1 on B-dataset and F-dataset. The experimental results are shown in Supplementary Figure S1. Regarding B-dataset, the AUC performance of KNN is gradually improved when the number of neighbors becomes larger, but the increase in AUC is much smaller when the number of neighbors is larger than 9. By applying the similar tuning process to the other classifiers, we could also obtain their parameter settings with the best performance. In particular, the hyperparameters of all classifiers as shown in Supplementary Table S3. The experimental results are presented in Table 3 and Figure 5. Besides, there are several points worth further commentary. First, among all classifiers, the performance of Gaussian NB is the worst. The main reason for its unsatisfactory performance is that Gaussian NB assumes the independence of features, which is difficult to be satisfied for the application of drug repositioning. Second, the performances of SVM and LR are fair, and thus the degree of nonlinearity in our datasets is yet to be verified. Third, although KNN is the second-best classifier, its ability of fault tolerance tends to become less efficient when the number of features increases. Lastly, as an efficient technique in ensemble learning, RF is preferred over the other classifiers due to its enhanced ability in processing high-dimensional data, which is the case of our datasets. As we know, there are many graph representation learning methods that can well learn the network representation of biomolecules in biological information networks. Moreover, the performance of GCN is moderate because of its excessive smoothness, and the difference in the performance between LINE and SDNE is rather small due to the fact that they share similar ideas of learning network representations for nodes. Since B-dataset and F-dataset are two different datasets, the promising performance of HINGRL on them could be, to some extent, an indicator to demonstrate its generalization ability. Rather than applying the HINGRL model trained on B-dataset to prediction the drug—disease associations in F-dataset, we adopt a different strategy by following \[ 45 , 46 \], which proposes to analyze the generalization ability on HINs with different sparsity by removing a certain proportion of drug—disease associations. The reason for this is due to the crucial constraint of DeepWalk, which requires DeepWalk to be retrained for learning the representations of new nodes in a given network \[ 47 \]. The main reason for that phenomenon is that the network representations of drugs and diseases can be enhanced by HINGRL if more heterogenous information about them are observed in training data. In summary, although the generalization ability of HINGRL is heavily dependent on the size of common drugs and diseases shared by two datasets used for training and testing respectively, the consideration of heterogenous information alleviates the effect resulted from the constraint of DeepWalk. To demonstrate the ability of HINGRL in discovering novel drug—disease associations, we have conducted additional experiments on the B-dataset. In particular, all known associations between drugs and diseases are used to compose the training dataset and then HINGRL is applied to verify unknown associations. An in-depth investigation into the experimental results is performed and several case studies are selected for further discussion as follows. As one of the drugs for the treatment of schizophrenia, clozapine has been deeply studied by many pharmacological scientists because of its remarkable clinical efficacy \[ 48 \]. In Table 7 , the top 10 disease candidates are predicted by HINGRL to have associations with clozapine, and 5 of them have already been experimentally confirmed by the relevant literature. In order to verify the rationality behind the prediction results, we take anxiety disorders as an example to explain why it is a potential disease that can be cured by clozapine in theory. As has been pointed out by \[ 49 \], anxiety disorders often occur as a common complication with schizophrenia due to the relationship between anxiety and the abnormal regulation of serotonin observed in the patients. Since clozapine can reduce the increase in serotonin caused by a noncompetitive antagonist of N-methyl-D-aspartate receptors \[ 50 \], we have reason to believe that clozapine is likely to produce a pharmacological effect for anxiety disorders. More evidences can be found in relevant databases. First, anxiety disorders and pain are two similar diseases as indicated by the DisGeNET database, and a known association between pain and clozapine has existed in the HIN of B-dataset. Second, according to the DrugBank database, the chemical structures of olanzapine and clozapine are similar as their cosine similarity is as large as 0. Breast neoplasms are the most common symptom in the female population. The top 10 candidates of potential drugs predicted by HINGRL are shown in Tables 8 and 6 of them have been recorded in literature to be effective when used to treat breast neoplasms. Cocaine obtains the largest prediction score among all unverified drugs, and an in-depth analysis is given after a systematic literature review. As indicated by \[ 51 \], celecoxib has an inhibitory effect on the growth of breast cancer cells containing cyclooxygenase-2, and it also has a verified association with breast neoplasms in B-dataset. According to the DrugBank database, celecoxib is associated with cocaine due to the fact that the combination of celecoxib and cocaine is able to slow down the metabolism of cells \[ 30 \]. In this regard, our findings indicate a possible treatment for breast neoplasms by the collaboration of celecoxib and cocaine. For HINGRL, its reasons regarding the discovery of the association between cocaine and breast neoplasms are 2-fold: 1 there are many neighboring nodes, i. Since more network paths are existed between them during random walk, the representations of cocaine and celecoxib are more similar from the perspective of network topology. Furthermore, the introduction of protein-related associations also strengthens the connectivity between cocaine and celecoxib. The similarity of attribute information between verified drugs and known drugs for breast neoplasms in B-dataset. The horizontal axis represents the known drugs, whereas the vertical axis represents the verified ones. The experimental results are presented in Figure 7 , and we note that each of the verified drugs is highly similar to some of the known drugs according to the distribution of blocks with dark color. In other words, HINGRL is able to identify these verified drugs for breast neoplasms solely from the perspective of biological knowledge. In sum, these case studies again demonstrate the promising accuracy of HINGRL in drug repositioning, and hence it is believed that HINGRL could be a useful tool to discover novel drug—disease associations especially for new diseases without any known associations. To capture the features of drugs and disease from a more comprehensive perspective, HINGRL first integrates protein-related associations and the biological knowledge of drugs and diseases into the original drug—diseases association network, thus composing a complicated HIN. After that, different graph representation learning techniques are utilized by HINGRL to capture the targeted features of drugs and diseases from the perspectives of network topology and biological knowledge. Experimental results on two benchmark datasets demonstrate that HINGRL yields a better performance than state-of-the-art drug repositioning algorithms in terms of accuracy and robustness. Our in-depth analysis of case study is also a strong indicator that HINGRL could be a useful tool to discover novel drug—disease associations especially for new diseases without any known associations. Regarding the future work, we would like to extend our research from four aspects. First, we are interested in exploring the possibility of applying HINGRL to other relevant applications, such as protein—protein interaction prediction \[ 52 , 53 \], and miRNA—disease association prediction \[ 54 \]. Second, regarding the construction of HIN, we intend to incorporate more specific information originated from the molecular mechanism of diseases and evaluate the importance of these heterogeneous information in drug repositioning. Last, since there are many other kinds of biological network, we aim to explore the possibility of proposing a better model that can adaptively learn the representations of drugs and diseases in a more complicated HIN. We integrate rich heterogeneous information, i. The authors would like to thank all anonymous reviewers for their constructive advice. Lun Hu received the B. His research interests include machine learning, complex network analytics and their applications in bioinformatics. Zhu-Hong You received his B. He obtained his Ph. His current research interests include neural networks, intelligent information processing, sparse representation, and its applications in bioinformatics. Lei Wang received the Ph. His research interests include data mining, pattern recognition, machine learning, deep learning, computational biology, and bioinformatics. Health Aff ; 25 : — 8. Google Scholar. Drug repositioning: identifying and developing new uses for existing drugs. Nat Rev Drug Discov ; 3 : — A survey of current trends in computational drug repositioning. Brief Bioinform ; 17 : 2 — Oral sildenafil in the treatment of erectile dysfunction. N Engl J Med ; : — A review of computational drug repositioning: strategies, approaches, opportunities, challenges, and directions. Biomedical data and computational models for drug repositioning: a comprehensive review. Brief Bioinform ; 22 2 : — Matrix factorization-based prediction of novel drug indications by integrating genomic space. Comput Math Methods Med ; Predicting potential side effects of drugs by recommender methods and ensemble learning. Neurocomputing ; : — Predicting drug-disease associations via multi-task learning based on collective matrix factorization. Front Bioeng Biotechnol ; 8 Computational drug repositioning using low-rank matrix approximation and randomized algorithms. Bioinformatics ; 34 : — Mol Syst Biol ; 7 : Drug repositioning by kernel-based integration of molecular structure, molecular activity, and phenotype data. PLoS One ; 8 :e Identification of drug-disease associations using information of molecular structures and clinical symptoms via deep convolutional neural network. Front Chem ; 7 : Predicting drug—disease associations through layer attention graph convolutional network. Brief Bioinform ; 22 4. Target identification among known drugs by deep learning from heterogeneous networks. Chem Sci ; 11 : — Drug repositioning based on comprehensive similarity measures and bi-random walk algorithm. Bioinformatics ; 32 : — A network integration approach for drug-target interaction prediction and computational drug repositioning from heterogeneous information. Nat Commun ; 8 : 1 — Bioinformatics ; 35 : — 8. Brief Bioinform ; 22 6. Computational drug repositioning based on multi-similarities bilinear matrix factorization. Hu L , Chan KC. Fuzzy clustering in a complex network based on content relevance and link structures. A variational Bayesian framework for cluster analysis in a complex network. HiSCF: leveraging higher-order structures for clustering analysis in biological networks. Bioinformatics ; 37 : — DTI-CDF: a cascade deep forest model towards the prediction of drug-target interactions based on hybrid features. Brief Bioinform ; 22 : — Comput Biol Med ; A survey on computational models for predicting protein—protein interactions. Brief Bioinform ; 22 5. Valproic acid, a histone deacetylase inhibitor, induces apoptosis in breast cancer stem cells. Chem Biol Interact ; : 51 — 8. The comparative toxicogenomics database: update Nucleic Acids Res ; 45 : D — 8. Predicting drug-disease associations by using similarity constrained matrix factorization. BMC Bioinformatics ; 19 : 1 — Drug Bank 5. Nucleic Acids Res ; 46 : D — DisGeNET: a comprehensive platform integrating information on human disease-associated genes and variants. Drug—drug similarity measure and its applications. Brief Bioinform ; A chemical language and information system. Introduction to methodology and encoding rules. J Chem Inf Comput Sci ; 28 : 31 — 6. Landrum G. Rdkit documentation. Release ; 1 : 1 — Predictive intelligence powered attentional stacking matrix factorization algorithm for the computational drug repositioning. Appl Soft Comput ; Cybernetics ;1—9. Autoencoder for words. Neurocomputing ; : 84 — Deepwalk: online learning of social representations. Extracting coevolutionary features from protein sequences for predicting protein-protein interactions. Kipf TN , Welling M. Semi-supervised classification with graph convolutional networks. Line: large-scale information network embedding. Google Preview. Structural deep network embedding. Grover A , Leskovec J. Briefings in Bioinformatics. Influence-aware graph neural networks. Applied Soft Computing ; Transl Psychiatry ; 11 : 1 — Anxiety disorders and schizophrenia. Curr Psychiatry Rep ; 6 : — Clozapine and haloperidol differently suppress the MKincreased glutamatergic and serotonergic transmission in the medial prefrontal cortex of the rat. Neuropsychopharmacology ; 32 : — Growth inhibition of breast cancer cells by celecoxib. Breast Cancer Res Treat ; 69 3 Identifying protein complexes from protein-protein interaction networks based on fuzzy clustering and GO semantic information. A distributed framework for large-scale protein-protein interaction data analysis and prediction using map reduce. Graph convolution for predicting associations between miRNA and drug resistance. Bioinformatics ; 36 : — 8. Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide. Sign In or Create an Account. Sign in through your institution. Advanced Search. Search Menu. Article Navigation. Close mobile search navigation Article Navigation. Volume Article Contents Abstract. Materials and methods. Results and discussion. Data availability. Journal Article. HINGRL: predicting drug—disease associations with graph representation learning on heterogeneous information networks. Oxford Academic. Lun Hu. Corresponding author. Zhu-Hong You. School of Computer Science. Lei Wang. Xiao-Rui Su. Revision received:. Select Format Select format. Permissions Icon Permissions. Abstract Identifying new indications for drugs plays an essential role at many phases of drug research and development. Figure 1. Open in new tab Download slide. Motivated by the observation that diseases are similar if the drugs they are associated with are also similar \[ 35 \], we extract the biological information of diseases in light of medical subject descriptors collected from the Medical Subject Headings MeSH thesaurus \[ 36 \]. In particular, the relationships among diseases are described by the MeSH tree structure and computed as representation vectors \[ 37 \]. To do so, each disease is first described with a directed acyclic graph DAG by the MeSH descriptors, and then the similarity between two diseases, i. When we incorporate autoencoder into HINGRL for dimension reduction, the biological information of drugs and diseases, i. The weight matrices, i. Unlike the biological information that only involves individual drugs and diseases, the network topology information observed in a HIN is more complicated, as it represents the relationship between pairwise nodes. DeepWalk takes pairwise nodes as input and learns the sequence representation of each node by following the random walk theory. The output of DeepWalk are the corresponding representation vectors of nodes obtained from a skip-gram model. Finally, a skip-gram model is adopted to calculate Equation 12 as indicated by the following equation. Table 1 Open in new tab. Experimental results of performance comparison on two benchmark datasets. B-dataset deepDR 0. Figure 2. Figure 3. Table 2 Open in new tab. Figure 4. Table 3 Open in new tab. Gaussian NB Figure 5. Table 4 Open in new tab. Figure 6. Table 5 Open in new tab. Table 6 Open in new tab. Table 7 Open in new tab. Evidence PMID. Clozapine Headache D 0. Table 8 Open in new tab. DrugBank ID. Breast neoplasms Valproic acid DB 0. Figure 7. Key Points. Google Scholar Crossref. Search ADS. Google Scholar PubMed. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals. Issue Section:. Download all slides. Supplementary data. Views 5, More metrics information. Total Views 5, Email alerts Article activity alert. Advance article alerts. New issue alert. In progress issue alert. Receive exclusive offers and updates from Oxford Academic. Citing articles via Web of Science Latest Most Read Most Cited A two-task predictor for discovering phase separation proteins and their undergoing mechanism. MetaDegron: multimodal feature-integrated protein language model for predicting E3 ligase targeted degrons. AptaDiff: de novo design and optimization of aptamers based on diffusion models. Statistical analysis of multiple regions-of-interest in multiplexed spatial proteomics data. Enhancing RNA-seq analysis by addressing all co-existing biases using a self-benchmarking approach with 2D structural insights. More from Oxford Academic. Bioinformatics and Computational Biology. Biological Sciences. Science and Mathematics. Authoring Open access Purchasing Institutional account management Rights and permissions. Get help with access Accessibility Contact us Advertising Media enquiries.

Urumqi where can I buy cocaine