Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) are closely related to the treatment of human diseases. Traditional biological experiments often require time-consuming and labor-intensive in their search for mechanisms of disease. Computational methods are regarded as an effective way to predict unknown lncRNA-miRNA interactions (LMIs). However, most of them complete their tasks by mainly focusing on a single lncRNA-miRNA network without considering the complex mechanism between biomolecular in life activities, which are believed to be useful for improving the accuracy of LMI prediction. To address this, a heterogeneous information network (HIN) learning model with neighborhood-level structural representation, called HINLMI, to precisely identify LMIs. In particular, HINLMI first constructs a HIN by integrating nine interactions of five biomolecules. After that, different representation learning strategies are applied to learn the biological and network representations of lncRNAs and miRNAs in the HIN from different perspectives. Finally, HINLMI incorporates the XGBoost classifier to predict unknown LMIs using final embeddings of lncRNAs and miRNAs. Experimental results show that HINLMI yields a best performance on the real dataset when compared with state-of-the-art computational models. Moreover, several analysis experiments indicate that the simultaneous consideration of biological knowledge and network topology of lncRNAs and miRNAs allows HINLMI to accurately predict LMIs from a more comprehensive perspective. The promising performance of HINLMI also reveals that the utilization of rich heterogeneous information can provide an alternative insight for HINLMI to identify novel interactions between lncRNAs and miRNAs.

Attributed graph clustering is a fundamental, yet challenging, task for studying underlying network structures. Recently, a variety of graph representation learning models have been proposed to effectively infer the node embeddings, which are then incorporated into conventional clustering techniques to identify meaningful clusters. While these models tend to preserve node proximities, which reflect the similarity between nodes in both structural and attribute dimensions, for representation learning, they generally overlook the crucial dependencies between node embeddings and the resulting clusters. To overcome this problem, we propose a novel fuzzy-based deep attributed graph clustering model, namely FDAGC, which is capable of achieving the task in a purely unsupervised and end-to-end manner without additionally incorporating conventional clustering techniques. In particular, FDAGC first encodes network structures and node attributes into a compact representation with graph convolution. A reconstruction error is then estimated to minimize the information loss during network message-passing. Besides, we utilize a self-monitoring training strategy to optimize node embeddings, thus improving the cluster cohesion by guiding them towards cluster centers. In the training phase, our expectations about resulting clusters are explicitly incorporated into the optimization of FDAGC via the concept of fuzzy clustering, thus leading to more accurate clustering by coupling the dependency between graph representation learning and attributed graph clustering. Extensive experiments have demonstrated the superior performance of FDAGC in terms of several evaluation metrics, such as Accuracy, Normalized Mutual Information, F1-score and Adjusted Rand Index, on six real-world attributed graphs with different scales.

Drug repositioning is a promising drug development technique to identify new indications for existing drugs. However, existing computational models only make use of lower-order biological information at the level of individual drugs, diseases and their associations, but few of them can take into account higher-order connectivity patterns presented in biological heterogeneous information networks (HINs). In this work, we propose a novel graph representation learning model, namely FuHLDR, for drug repositioning by fusing higher and lower-order biological information. Specifically, given a HIN, FuHLDR first learns the representations of drugs and diseases at a lower-order level by considering their biological attributes and drug-disease associations (DDAs) through a graph convolutional network model. Then, a meta-path-based strategy is designed to obtain their higher-order representations involving the associations among drugs, proteins and diseases. Their integrated representations are thus determined by fusing higher and lower-order representations, and finally a Random Vector Functional Link Network is employed by FuHLDR to identify novel DDAs. Experimental results on two benchmark datasets demonstrate that FuHLDR performs better than several state-of-the-art drug repositioning models. Furthermore, our case studies on Alzheimer's disease and Breast neoplasms indicate that the rich higher-order biological information gains new insight into drug repositioning with improved accuracy.

Protein-protein interactions (PPIs) play a critical role in the proteomics study, and a variety of computational algorithms have been developed to predict PPIs. Though effective, their performance is constrained by high false-positive and false-negative rates observed in PPI data. To overcome this problem, a novel PPI prediction algorithm, namely PASNVGA, is proposed in this work by combining the sequence and network information of proteins via variational graph autoencoder. To do so, PASNVGA first applies different strategies to extract the features of proteins from their sequence and network information, and obtains a more compact form of these features using principal component analysis. In addition, PASNVGA designs a scoring function to measure the higher-order connectivity between proteins and so as to obtain a higher-order adjacency matrix. With all these features and adjacency matrices, PASNVGA trains a variational graph autoencoder model to further learn the integrated embeddings of proteins. The prediction task is then completed by using a simple feedforward neural network. Extensive experiments have been conducted on five PPI datasets collected from different species. Compared with several state-of-the-art algorithms, PASNVGA has been demonstrated as a promising PPI prediction algorithm. The source code of PASNVGA and all datasets are available at https://github.com/weizhi-code/PASNVGA .

Attributed graph clustering is of significance for an in-depth understanding of the intrinsic organization of complex networks. Recently, owing to the powerful learning capability of deep neural networks, increasing attention has been paid to developing more advanced deep learning approaches for attributed graph clustering. However, these approaches mainly focus on the purpose of obtaining general node representations, followed by clustering as a downstream task in combination with traditional techniques. Intuitively, the clustering performance is further limited due to the lack of achieving the desirable clustering quality in the graph representation learning process. To this end, this paper proposes a novel end-to-end attributed graph clustering model, namely FCGCN, by integrating fuzzy clustering and graph convolution network. FCGCN is trained toward optimizing an unsupervised fuzzy-based clustering objective, which is specifically formulated for precisely evaluating the clustering quality by considering the fuzzy memberships of nodes over clusters. To avoid the generation of undesirable clusters, we introduce a tailored regularization term by alleviating the over-smoothing issue on graph neural networks. By doing so, an explicit connection between graph representation learning and clustering can thus be established, considerably improving the clustering performance. To evaluate the performance of FCGCN, extensive experiments are conducted on a total of five real attributed graph datasets, and the results demonstrate the superiority of FCGCN over several state-of-the-art algorithms in terms of effectiveness and efficiency.

To understand the mechanisms of complex systems, attributed graphs (AGs) are recognized as a valuable model by their capability of describing nontrivial topological structures and rich node contents, and their emergence raises new challenges on the task of graph clustering. Although a variety of computational algorithms have been proposed to perform accurate clustering analysis on AGs, most of them are incapable of inferring the cluster labels of nodes through links, thus falling short of explaining node behaviors on how to formulate overlapping clusters. Moreover, the vast amount of links considerably decreases the computation efficiency if they are explicitly taken into account for AG clustering. To overcome this problem, we present a novel variational Bayesian learning model, which avoids generating a complete AG by only simulating the generative process of its skeleton with the prior knowledge on the cluster labels of links. When addressing the inference problem, we develop an efficient algorithm, namely, LCAAG, for determining the optimal cluster labels of nodes by estimating local community structures of links. The convergence of LCAAG has been proved theoretically. Compared with several state-of-the-art algorithms, LCAAG has demonstrated its promising performance in terms of both accuracy and scalability on five different scaled benchmark datasets. The source code and datasets are available at https://github.com/shallowdreamoon/LCAAG.git.