Abstract Self-supervised learning (SSL) has become a promising and popular learning paradigm for graph data, offering the advantage of capturing informative knowledge without reliance on manual labels. As a representative class of generative graph SSL models, existing graph autoencoders (GAE) excel in link prediction tasks and are steadily improving in node classification tasks. However, GAE is essentially based on the Information Maximization (InfoMax) principle, always captures much redundant information. In this paper, we propose an Alternating-update-strategy based Graph Autoencoder, including alternating update module (AUM) and GAE. For AUM, we design an Alternating-update-strategy to generate a new graph with reduced redundancy, in order to reduce the amount of redundant information that the encoder may capture. For GAE, we feed it the new graph and employ a re-mask decoding strategy to generate node representations. Our model is evaluated on five common real-world datasets for the node classification task, and the experimental results demonstrate its superiority. Meanwhile, our model has also achieved excellent results in specific e-commerce warehousing application scenarios.

Graph neural networks (GNNs) are capable of modeling graph data using various types of nodes and edges, and thus can be widely used in the fields of recommender systems and bioinformatics. However, most existing graph neural network models do not alleviate the problem of incomplete raw graph data and the need for extensive labels. This leads to increased labor costs and produces suboptimal or even incorrect results. In this article, we propose an end-to-end method for fusing learnable node-level and edge-level augmentation in automatic graph-level contrastive learning (NEAGCL), consisting of a graph preprocessing module and tailored graph-level contrastive learning. In the graph preprocessing module, we propose a parallel view generator method, which solves the problem of data incompleteness by simultaneously performing adaptive structure enhancement and learnable node information optimization. In tailored graph-level contrastive learning, we employ contrast learning in self-supervised learning and classification loss to jointly train graph classifiers, thus solving the problem of relying on the original labels. Besides, our model employs differentiable contrastive learning and is an end-to-end self-supervised model. Comparative tests of semisupervised learning and unsupervised learning on seven benchmark datasets for graph-level tasks demonstrate the superiority of our model. We demonstrate that effective graph contrastive learning requires learning better joint representations of graph structures and node attributes to improve the performance of graph-level tasks.

The off-target activities within the CRISPR-Cas9 system remains a formidable barrier to its broader application and development. Recent advancements have highlighted the potential of deep learning models in predicting these off-target effects, yet they encounter significant hurdles including imbalances within datasets and the intricacies associated with encoding schemes and model architectures. To surmount these challenges, our study innovatively introduces an Efficiency and Specificity-Based (ESB) class rebalancing strategy, specifically devised for datasets featuring mismatches-only off-target instances, marking a pioneering approach in this realm. Furthermore, through a meticulous evaluation of various One-hot encoding schemes alongside numerous hybrid neural network models, we discern that encoding and models of moderate complexity ideally balance performance and efficiency. On this foundation, we advance a novel hybrid model, the CRISPR-MCA, which capitalizes on multi-feature extraction to enhance predictive accuracy. The empirical results affirm that the ESB class rebalancing strategy surpasses five conventional methods in addressing extreme dataset imbalances, demonstrating superior efficacy and broader applicability across diverse models. Notably, the CRISPR-MCA model excels in off-target effect prediction across four distinct mismatches-only datasets and significantly outperforms contemporary state-of-the-art models in datasets comprising both mismatches and indels. In summation, the CRISPR-MCA model, coupled with the ESB rebalancing strategy, offers profound insights and a robust framework for future explorations in this field.

Heterogeneous information network (HIN) has been applied in a wide variety of graph analysis tasks. At present, it is a trend of heterogeneous graph neural networks (HGNNs) to cast the meta-paths aside, since it solves the problem of structural information loss caused by artificially designed meta-paths. However, existing meta-path-free HGNNs fail to take into account that most node types in many HINs have no attributes, and they cannot make full use of sparse node attributes when applied to HINs with missing attributes. Furthermore, their computation of attention coefficients explores the correlations of node attributes while almost ignoring structural ones, which may limit the expression ability of the model and cause overfitting in model training. To alleviate these issues, we propose an HGNN with attribute enhancement and structure-aware attention (HGNN-AESA). First, we design an attribute enhancement module (AEM) to connect more useful attributed nodes to the target nodes. Specifically, AEM introduces a random walk with restart (RWR) strategy to obtain structural relevance scores of each node within its specific subgraph. The structural relevance scores are used to capture potentially influential attributed nodes in high-order neighborhood for each target node. Second, we propose heterogeneous structure-aware attention layers (HSALs) to learn node representations. HSALs follow a hierarchical attention framework, including node-level and type-level attention. The node-level attention aggregates feature (attribute) embeddings of same-type neighbors, and the relevant attention coefficients depend on the combination of node attributes and heterogeneous structural interventions. The type-level attention fuses all type-specific vector representations and generates the ultimate node embedding. Finally, extensive experiments on three different real-world HIN datasets demonstrate that our model outperforms state-of-the-art methods.

The continuous interaction of users and information aggregation has become a social phenomena over massive social media platforms. However, the uncertainty of users’ behavior is leading great challenges to social networks analysis in terms of system structure, evolution characteristics, dynamic behavior, and so forth. Thus, this article proposes a formal user behavior modeling and analysis approach. First, aiming at identifying the behavior patterns of user activity sequence, we present a user activity transition system model based on stochastic Petri net (SPN), which can formally depict the process and structures of social users click activities. Then, the average number of tokens in each place, the probability density function of the tokens, the token flow rate of transitions, and the time spent in each state are analyzed by isomorphic it into a Markov chain (MC), respectively. These four indicators are used to evaluate the performance of the proposed system model. The experimental results demonstrate that the proposed approach can help us to understand the rules of users’ first activity and activity preferences, so as to provide practical suggestions for the development of social networking platforms and content recommendation.

Numerous experimental studies have indicated that alteration and dysregulation in mircroRNAs (miRNAs) are associated with serious diseases. Identifying disease-related miRNAs is therefore an essential and challenging task in bioinformatics research. Computational methods are an efficient and economical alternative to conventional biomedical studies and can reveal underlying miRNA-disease associations for subsequent experimental confirmation with reasonable confidence. Despite the success of existing computational approaches, most of them only rely on the known miRNA-disease associations to predict associations without adding other data to increase the prediction accuracy, and they are affected by issues of data sparsity. In this paper, we present MRRN, a model that combines matrix reconstruction with node reliability to predict probable miRNA-disease associations. In MRRN, the most reliable neighbors of miRNA and disease are used to update the original miRNA-disease association matrix, which significantly reduces data sparsity. Unknown miRNA-disease associations are reconstructed by aggregating the most reliable first-order neighbors to increase prediction accuracy by representing the local and global structure of the heterogeneous network. Five-fold cross-validation of MRRN produced an area under the curve (AUC) of 0.9355 and area under the precision-recall curve (AUPR) of 0.2646, values that were greater than those produced by comparable models. Two different types of case studies using three diseases were conducted to demonstrate the accuracy of MRRN, and all top 30 predicted miRNAs were verified.

Location recommendation method is an important application in a location-based social network. At present, it is a trend to integrate different recommendation methods since they have their own advantages in capturing different preferences of users and an integrated method can generally provide a better performance than every individual. However, the existing integration policies do not learn user preferences in their integration processes so that they cannot make full use of the advantage of each method. Therefore, we propose a novel integration method: learning-to-rank-based integration. In our method, a confidence coefficient is applied for each user in the integration process, and these coefficients can well optimize recommendation performance. A learning-to-rank-based algorithm is designed to train the confidence coefficients. A group of experiments are done on a real large-scale check-in data set, and the results demonstrate that our method outperforms the state-of-the-art ones.

It is a valuable study for Location-based Social Network (LBSN) make a more accurate Points-of-Interest (POI) recommendation since that can improve users' experiences. There have been many methods of POIs recommendation that consider context, personal preference pattern, and/or matrix factorization. However, the continuous contexts have not been thoroughly considered in these methods. This paper first proposes a locations splitting method which can handle both continuous and discrete contexts. Moreover, we present a Context-aware Probabilistic Matrix Factorization method (CPMF) that factorizes a frequency matrix of contexts and locations to obtain the user-location checkin probabilities. We design a Personal Preference Confidence (PPC) to extract a set of reliable POIs with confidence values for every user. Finally, we propose a hybrid recommender which fuses CPMF with PPC to recommend top-n POIs. Experiments on a large-scale real-world checkins dataset demonstrate that our recommendation method obtains a well performance and effect.