Graph Embedding Methods Research Articles

Graphfingerprint: graph embedding of graphs with almost constant sub-structures

AbstractIn some machine learning applications, graphs tend to be composed of a large number of tiny almost constant sub-structures. The current embedding methods are not prepared for this type of graphs and thus, their representational power tends to be very low. Our aim is to define a new graph embedding that considers this specific type of graphs. We present GraphFingerprint, which is a new embedding method that specifically considers the fact that graphs are composed of millions of almost constant sub-structures. The three-dimensional characterisation of a chemical metal-oxide nanocompound easily fits in these types of graphs, which nodes are atoms and edges are their bonds. Our graph embedding method has been used to predict the toxicity of these nanocompounds, achieving a high accuracy compared to other embedding methods. The representational power of the current embedding methods do not properly satisfy the requirements of some machine learning applications based on graphs, for this reason, a new embedding method has been defined and heuristically demonstrated that achieves good accuracy.

Large-Scale Knowledge Graphs as a Tool for Enhanced Robotic Perception

Autonomous robotic systems depend on their perception and understanding of their environment for informed decision-making. One of the goals of the Semantic Web is to make knowledge on the Web machine-readable, which can significantly aid robots by providing background knowledge, and thereby support their understanding. In this paper, we present a reasoning system that uses the Ontology for Robotic Knowledge Acquisition (ORKA) to integrate the sensory data and perception algorithms of the robot, thereby enhancing its autonomous capabilities. This reasoning system is subsequently employed to retrieve and integrate information from the Semantic Web, thereby improving the robot's comprehension of its environment. To achieve this, the system employs a Perceived-Entity Linking (PEL) pipeline that associates regions in the sensory data of the robotic agent with concepts in a target knowledge graph. As a use-case for the linking process, the Perceived-Entity Typing task is used to determine the more fine-grained subclass of the perceived entities. Specifically, we provide an analysis of the performance of different knowledge graph embedding methods on the task using a annotated observations and WikiData as a target knowledge graph. The experiments indicate that relying on pre-trained embedding methods results in an increased performance when using TransE as the embedding method for the observations of the robot. This contribution advances the field by demonstrating the potential of integrating Semantic Web technologies with robotic perception, thereby enabling more nuanced and context-aware decision-making in autonomous systems.

Network community detection via neural embeddings

Recent advances in machine learning research have produced powerful neural graph embedding methods, which learn useful, low-dimensional vector representations of network data. These neural methods for graph embedding excel in graph machine learning tasks and are now widely adopted. However, how and why these methods work—particularly how network structure gets encoded in the embedding—remain largely unexplained. Here, we show that node2vec—shallow, linear neural network—encodes communities into separable clusters better than random partitioning down to the information-theoretic detectability limit for the stochastic block models. We show that this is due to the equivalence between the embedding learned by node2vec and the spectral embedding via the eigenvectors of the symmetric normalized Laplacian matrix. Numerical simulations demonstrate that node2vec is capable of learning communities on sparse graphs generated by the stochastic blockmodel, as well as on sparse degree-heterogeneous networks. Our results highlight the features of graph neural networks that enable them to separate communities in the embedding space.

Joint embedding in Hierarchical distance and semantic representation learning for link prediction

The link prediction task aims to predict missing entities or relations in the knowledge graph and is essential for the downstream application. Existing well-known models deal with this task by mainly focusing on representing knowledge graph triplets in the distance space or semantic space. However, they can not fully capture the information of head and tail entities, nor even make good use of hierarchical level information. Thus, in this paper, we propose a novel knowledge graph embedding model for the link prediction task, namely, HIE, which models each triplet (h, r, t) into distance measurement space and semantic measurement space, simultaneously. Moreover, HIE is introduced into hierarchical-aware space to leverage rich hierarchical information of entities and relations for better representation learning. Specifically, we apply distance transformation operation on the head entity in distance space to obtain the tail entity instead of translation-based or rotation-based approaches. Experimental results of HIE on four real-world datasets show that HIE outperforms several existing state-of-the-art knowledge graph embedding methods on the link prediction task and deals with complex relations accurately.

VulTR: Software vulnerability detection model based on multi-layer key feature enhancement

Software vulnerabilities pose a huge threat to current network security, which continues to lead to data leaks and system damage. In order to effectively identify and patch these vulnerabilities, researchers have proposed automated detection methods based on deep learning. However, most of the existing methods only rely on single-dimensional data representation and fail to fully explore the composite characteristics of the code. Among them, the sequence embedding method fails to effectively capture the structural characteristics of the code, while the graph embedding method focuses more on the global characteristics of the overall graph structure and is still insufficient in optimizing the representation of nodes. In view of this, this paper constructs the VulTR model, which incorporates an importance assessment mechanism to strengthen the key syntax levels of the source code (from lexical elements to nodes and graph-level structures), significantly improving the importance of key vulnerability features in classification decisions. At the same time, a relationship connection diagram is constructed to describe the spatial characteristics of the correlations between functions. Experimentally verified, VulTR's F1 scores on both synthetic and real data sets exceed those of the compared models (VulDeePecker, SySeVR, Devign, VulCNN, IVDetect, and mVulPreter).

FuseLinker: Leveraging LLM’s pre-trained text embeddings and domain knowledge to enhance GNN-based link prediction on biomedical knowledge graphs

ObjectiveTo develop the FuseLinker, a novel link prediction framework for biomedical knowledge graphs (BKGs), which fully exploits the graph’s structural, textual and domain knowledge information. We evaluated the utility of FuseLinker in the graph-based drug repurposing task through detailed case studies. MethodsFuseLinker leverages fused pre-trained text embedding and domain knowledge embedding to enhance the graph neural network (GNN)-based link prediction model tailored for BKGs. This framework includes three parts: a) obtain text embeddings for BKGs using embedding-visible large language models (LLMs), b) learn the representations of medical ontology as domain knowledge information by employing the Poincaré graph embedding method, and c) fuse these embeddings and further learn the graph structure representations of BKGs by applying a GNN-based link prediction model. We evaluated FuseLinker against traditional knowledge graph embedding models and a conventional GNN-based link prediction model across four public BKG datasets. Additionally, we examined the impact of using different embedding-visible LLMs on FuseLinker’s performance. Finally, we investigated FuseLinker’s ability to generate medical hypotheses through two drug repurposing case studies for Sorafenib and Parkinson’s disease. ResultsBy comparing FuseLinker with baseline models on four BKGs, our method demonstrates superior performance. The Mean Reciprocal Rank (MRR) and Area Under receiver operating characteristic Curve (AUROC) for KEGG50k, Hetionet, SuppKG and ADInt are 0.969 and 0.987, 0.548 and 0.903, 0.739 and 0.928, and 0.831 and 0.890, respectively. ConclusionOur study demonstrates that FuseLinker is an effective novel link prediction framework that integrates multiple graph information and shows significant potential for practical applications in biomedical and clinical tasks. Source code and data are available at https://github.com/YKXia0/FuseLinker.

Community knowledge graph abstraction for enhanced link prediction: A study on PubMed knowledge graph

Objective:As new knowledge is produced at a rapid pace in the biomedical field, existing biomedical Knowledge Graphs (KGs) cannot be manually updated in a timely manner. Previous work in Natural Language Processing (NLP) has leveraged link prediction to infer the missing knowledge in general-purpose KGs. Inspired by this, we propose to apply link prediction to existing biomedical KGs to infer missing knowledge. Although Knowledge Graph Embedding (KGE) methods are effective in link prediction tasks, they are less capable of capturing relations between communities of entities with specific attributes (Fanourakis et al., 2023). Methods:To address this challenge, we proposed an entity distance-based method for abstracting a Community Knowledge Graph (CKG) from a simplified version of the pre-existing PubMed Knowledge Graph (PKG) (Xu et al., 2020). For link prediction on the abstracted CKG, we proposed an extension approach for the existing KGE models by linking the information in the PKG to the abstracted CKG. The applicability of this extension was proved by employing six well-known KGE models: TransE, TransH, DistMult, ComplEx, SimplE, and RotatE. Evaluation metrics including Mean Rank (MR), Mean Reciprocal Rank (MRR), and Hits@k were used to assess the link prediction performance. In addition, we presented a backtracking process that traces the results of CKG link prediction back to the PKG scale for further comparison. Results:Six different CKGs were abstracted from the PKG by using embeddings of the six KGE methods. The results of link prediction in these abstracted CKGs indicate that our proposed extension can improve the existing KGE methods, achieving a top-10 accuracy of 0.69 compared to 0.5 for TransE, 0.7 compared to 0.54 for TransH, 0.67 compared to 0.6 for DistMult, 0.73 compared to 0.57 for ComplEx, 0.73 compared to 0.63 for SimplE, and 0.85 compared to 0.76 for RotatE on their CKGs, respectively. These improved performances also highlight the wide applicability of the extension approach. Conclusion:This study proposed novel insights into abstracting CKGs from the PKG. The extension approach indicated enhanced performance of the existing KGE methods and has applicability. As an interesting future extension, we plan to conduct link prediction for entities that are newly introduced to the PKG.

An open-set recognition method for ship radiated noise signal based on graph convolutional neural network prototype learning

The underwater acoustic perception system often undertakes multi-class ship recognition tasks. As the underwater acoustic recognition environment becomes increasingly complex, underwater acoustic perception systems often face various interference, such as unknown ships radiated noise signals and decoy signals. This poses higher requirements on the robustness and recognition capabilities of underwater acoustic target recognition methods. In this paper, we transform the problem of underwater acoustic target recognition in this scenario into an open-set recognition problem. We design a deep learning model based on graph convolutional neural networks, propose a graph embedding method for time-domain ships radiated noise signals, and propose a fully parameterized prototype learning framework. We simulate decoy signals in real sea areas, and all the data used in the experiments are actual collected data. The fully parameterized prototype learning framework based on a data-driven approach can not only effectively resist interference from unknown ship-radiated noise signals and decoy signals, but also accurately identify multi-class target ship-radiated noise signals. Ultimately, our method achieves end-to-end open-set recognition of ship-radiated noise signals.

Finding Efficient Graph Embeddings and Processing them by a CNN-based Tool

We introduce new tools to support finding efficient graph embedding techniques for graph databases and to process their outputs using deep learning for classification scenarios. Accordingly, we investigate the possibility of creating an ensemble of different graph embedding methods to raise accuracy and present an interconnected neural network-based ensemble to increase the efficiency of the member classification algorithms. We also introduce a new convolutional neural network-based architecture that can be generally proposed to process vectorized graph data provided by various graph embedding methods and compare it with other architectures in the literature to show the competitiveness of our approach. We also exhibit a statistical-based inhomogeneity level estimation procedure to select the optimal embedding for a given graph database efficiently. The efficiency of our framework is exhaustively tested using several publicly available graph datasets and numerous state-of-the-art graph embedding techniques. Our experimental results for classification tasks have proved the competitiveness of our approach by outperforming the state-of-the-art frameworks.

MARML: Motif-Aware Deep Representation Learning in Multilayer Networks.

The rapid increase in high-throughput, complex, and heterogeneous data has led to the adoption of network-structured models and analyses for interpretation. However, these data are inherently complex and challenging to understand, prompting researchers to turn to graph embedding methods to facilitate analysis. While general network embedding techniques have shown promise in improving downstream prediction and classification tasks, real-world data are complicated due to cross-domain interactions between different types of entities. Multilayered networks have been successful in integrating biological data to represent biological systems' hierarchy, but embedding nodes based on different types of interactions remains an unsolved problem. To address this challenge, we propose the Motif-aware deep representation learning in multilayer (MARML) networks for learning network representations. Our method considers recurring motif patterns, topological information, and attributive information from other sources as node features. We validated the MARML method using various multilayer network datasets. In addition, by incorporating motif information, MARML considers higher order connections across different hierarchies. The learned features exhibited excellent accuracy in tasks related to link prediction and link differentiation, enabling us to distinguish between existing and disconnected triplets. Through the integration of both intrinsic node attributes and topological network structures, we enhance our understanding of complex biological systems.

Predicting drug-target interactions by measuring confidence with consistent causal neighborhood interventions

Predicting drug-target interactions (DTI) is a crucial stage in drug discovery and development. Understanding the interaction between drugs and targets is essential for pinpointing the specific relationship between drug molecules and targets, akin to solving a link prediction problem using information technology. While knowledge graph (KG) and knowledge graph embedding (KGE) methods have been rapid advancements and demonstrated impressive performance in drug discovery, they often lack authenticity and accuracy in identifying DTI. This leads to increased misjudgment rates and reduced efficiency in drug development. To address these challenges, our focus lies in refining the accuracy of DTI prediction models through KGE, with a specific emphasis on causal intervention confidence measures (CI). These measures aim to assess triplet scores, enhancing the precision of the predictions. Comparative experiments conducted on three datasets and utilizing 9 KGE models reveal that our proposed confidence measure approach via causal intervention, significantly improves the accuracy of DTI link prediction compared to traditional approaches. Furthermore, our experimental analysis delves deeper into the embedding of intervention values, offering valuable insights for guiding the design and development of subsequent drug development experiments. As a result, our predicted outcomes serve as valuable guidance in the pursuit of more efficient drug development processes.

Urban streets profiling with coupled spatio-temporal characteristics and topological information from the biking perspective

Urban street profiling is the spatio-temporal pattern discovery of street-level urban areas, which plays a vital role in understanding urban structures and dynamics. Due to the natural topology and various geographic characteristics on the streets, it is necessary to combine multi-dimensional spatio-temporal information to understand different profiles of streets. This research aims to develop a street profiling framework according to the coupled characteristics of streets. At the start, a bidirected dual graph and a spatial weighted graph embedding method were used to solve the street representation. Then, the street profiles can be extracted by clustering embedding vectors of streets and feature importance analysis. As the case study, we employed the bike trajectories and street view images in Xiamen, China to depict the geographic attributes of streets. The results can reveal nine spatio-temporal street profiles from the biking perspective, including three spatial distribution patterns and two spatial semantic patterns. Urban streets in the study area show a significant hierarchical pattern because of locations and the spatial lags of the biking behaviors. Meanwhile, the spatio-temporal characteristics of biking behaviors are the main factors of street profiles, though the street environment attributes participate in over half the number of profile types. We further evaluated the profiling ability of the proposed framework and the importance of urban street profiles using coupled characteristics. Overall, this study explored the profiling method for coupling static and dynamic characteristics of urban streets. The profiling results also help understand street usage and experiences by bikers, which have a practical value on the human-oriented classification of streets and further urban development from a geographic view.

DualNetGO: a dual network model for protein function prediction via effective feature selection.

Protein-protein interaction (PPI) networks are crucial for automatically annotating protein functions. As multiple PPI networks exist for the same set of proteins that capture properties from different aspects, it is a challenging task to effectively utilize these heterogeneous networks. Recently, several deep learning models have combined PPI networks from all evidence, or concatenated all graph embeddings for protein function prediction. However, the lack of a judicious selection procedure prevents the effective harness of information from different PPI networks, as these networks vary in densities, structures, and noise levels. Consequently, combining protein features indiscriminately could increase the noise level, leading to decreased model performance. We develop DualNetGO, a dual-network model comprised of a Classifier and a Selector, to predict protein functions by effectively selecting features from different sources including graph embeddings of PPI networks, protein domain, and subcellular location information. Evaluation of DualNetGO on human and mouse datasets in comparison with other network-based models shows at least 4.5%, 6.2%, and 14.2% improvement on Fmax in BP, MF, and CC gene ontology categories, respectively, for human, and 3.3%, 10.6%, and 7.7% improvement on Fmax for mouse. We demonstrate the generalization capability of our model by training and testing on the CAFA3 data, and show its versatility by incorporating Esm2 embeddings. We further show that our model is insensitive to the choice of graph embedding method and is time- and memory-saving. These results demonstrate that combining a subset of features including PPI networks and protein attributes selected by our model is more effective in utilizing PPI network information than only using one kind of or concatenating graph embeddings from all kinds of PPI networks. The source code of DualNetGO and some of the experiment data are available at: https://github.com/georgedashen/DualNetGO.

Spatio-temporal knowledge embedding method considering the lifecycle of geographical entities

With the emergence of substantial amounts of data featuring spatio-temporal characteristics, research on spatio-temporal knowledge has garnered widespread attention. However, the existing knowledge graph embedding methods are difficult to embed spatio-temporal knowledge in continuous time, failing to meet the unique embedding requirements of geographic entities for uninterrupted, constantly changing, and smoothly continuous embeddings. To address this issue, we begin by outlining a version-based representation of spatio-temporal knowledge. Building on this foundation, we propose a novel and effective knowledge graph embedding model called Version Embedding (VerE), which embeds continuous time into vector space, transforms geographic entities into corresponding versions using an attention mechanism, and calculates spatio-temporal knowledge scores under the constraint of version similarity regularization. Subsequently, we introduce a method for link prediction based on unknown time with the aim of assessing the model’s generalization capabilities in time representation. Finally, experiments conducted on two real-world datasets demonstrate significant performance improvement of VerE compared to most existing models, and its ability to maintain stability and continuity in link prediction under unknown time. These experiments confirmed the effectiveness of the proposed model and provided new perspectives and methods for spatio-temporal knowledge embedding.

Autobalanced Multitask Node Embedding Framework for Intelligent Education.

Recently, online education has become popular. Many e-learning platforms have been launched with various intelligent services aimed at improving the learning efficiency and effectiveness of learners. Graphs are used to describe the pairwise relations between entities, and the node embedding technique is the foundation of many intelligent services, which have received increasing attention from researchers. However, the graph in the intelligent education scenario has three noteworthy properties, namely, heterogeneity, evolution, and lopsidedness, which makes it challenging to implement ecumenical node embedding methods on it. In this article, an autobalanced multitask node embedding model is proposed, named MNE, and applied to the interaction graph, settling a few actual tasks in intelligent education. More specifically, MNE builds two purpose-built self-supervised node embedding learning tasks for heterogeneous evolutive graphs. Edge-specific reconstruction tasks are built according to the semantic information and properties of the heterogeneous edges, and an evolutive weight regression task is designed, aiding the model to perceive the evolution of learners' implicit cognitive states. Then, both aleatoric and epistemic uncertainty quantification techniques are introduced, achieving both task- and node-level weight estimation and instructing subtask autobalancing. Experimental results on real-world datasets indicate that the proposed model outperforms the state-of-the-art graph embedding methods on two assessment tasks and demonstrates the validity of the proposed multitask framework and subtask balancing mechanism. Our implementations are available at https://github.com/ccnu-mathits/MNE4HEN.

DCNeT: A disease comorbidity network-based temporal deep learning framework to predict cardiovascular risk in patients with mental disorders

Patients with mental disorders (MDs) are at higher subsequent risk of developing cardiovascular diseases (CVDs) than the general population. Early identification of cardiovascular risk in patients with MDs is beneficial for timely intervention and reducing disease burden. Recently, deep learning approaches have been increasingly applied in CVDs risk prediction. However, these methods have three major issues: 1) mostly relying on multiple types of clinical data, 2) not sufficiently mining and utilizing comorbidity patterns hidden in complex correlations among various diseases, and 3) not fully leveraging the time information, including the irregular intervals. To address these issues, we propose a disease comorbidity network-based temporal deep learning framework (DCNeT) to predict the subsequent CVDs risk for patients with MDs based on routinely collected administrative health data. Firstly, to identify the comorbidity patterns of MDs, we construct a disease comorbidity network (DCN) for MDs and apply graph embedding method to generate disease embeddings for each disease in the DCN. Then, a code attention mechanism is proposed to obtain the weight of each disease which is embedded into dense vectors based on the structure of DCN. We present a view attention mechanism to compute the attention weights of different types of features including disease embeddings, basic features, and disease indicators for generating the final representations of patients’ hospitalizations. Furthermore, to fully utilize the information on the irregular time intervals between hospitalizations, a time encoding module is designed, and the time-aware LSTM is adopted to model the irregular time intervals and capture the temporal patterns of patients’ hospitalizations. The experimental results show that DCNeT outperforms the state-of-the-art methods, with the area under the receiver operating characteristic curve of 0.7658, 0.8143, 0.8110, and 0.7839 on four datasets, respectively. The ablation experiments further demonstrate that each module of DCNeT, including the code attention, view attention, and time encoding module, contributes to its superior performance, with average improvements of 1.20 %, 1.65 %, and 1.13 % in accuracy, respectively. Our DCNeT could be utilized as an efficient framework for identifying high-risk groups of CVDs among patients with MDs that may benefit from screening and preventive strategies.

The RDF2vec family of knowledge graph embedding methods

Knowledge graph embeddings represent a group of machine learning techniques which project entities and relations of a knowledge graph to continuous vector spaces. RDF2vec is a scalable embedding approach rooted in the combination of random walks with a language model. It has been successfully used in various applications. Recently, multiple variants to the RDF2vec approach have been proposed, introducing variations both on the walk generation and on the language modeling side. The combination of those different approaches has lead to an increasing family of RDF2vec variants. In this paper, we evaluate a total of twelve RDF2vec variants on a comprehensive set of benchmark models, and compare them to seven existing knowledge graph embedding methods from the family of link prediction approaches. Besides the established GEval benchmark introducing various downstream machine learning tasks on the DBpedia knowledge graph, we also use the new DLCC (Description Logic Class Constructors) benchmark consisting of two gold standards, one based on DBpedia, and one based on synthetically generated graphs. The latter allows for analyzing which ontological patterns in a knowledge graph can actually be learned by different embedding. With this evaluation, we observe that certain tailored RDF2vec variants can lead to improved performance on different downstream tasks, given the nature of the underlying problem, and that they, in particular, have a different behavior in modeling similarity and relatedness. The findings can be used to provide guidance in selecting a particular RDF2vec method for a given task.

BJLD-CMI: a predictive circRNA-miRNA interactions model combining multi-angle feature information.

Increasing research findings suggest that circular RNA (circRNA) exerts a crucial function in the pathogenesis of complex human diseases by binding to miRNA. Identifying their potential interactions is of paramount importance for the diagnosis and treatment of diseases. However, long cycles, small scales, and time-consuming processes characterize previous biological wet experiments. Consequently, the use of an efficient computational model to forecast the interactions between circRNA and miRNA is gradually becoming mainstream. In this study, we present a new prediction model named BJLD-CMI. The model extracts circRNA sequence features and miRNA sequence features by applying Jaccard and Bert's method and organically integrates them to obtain CMI attribute features, and then uses the graph embedding method Line to extract CMI behavioral features based on the known circRNA-miRNA correlation graph information. And then we predict the potential circRNA-miRNA interactions by fusing the multi-angle feature information such as attribute and behavior through Autoencoder in Autoencoder Networks. BJLD-CMI attained 94.95% and 90.69% of the area under the ROC curve on the CMI-9589 and CMI-9905 datasets. When compared with existing models, the results indicate that BJLD-CMI exhibits the best overall competence. During the case study experiment, we conducted a PubMed literature search to confirm that out of the top 10 predicted CMIs, seven pairs did indeed exist. These results suggest that BJLD-CMI is an effective method for predicting interactions between circRNAs and miRNAs. It provides a valuable candidate for biological wet experiments and can reduce the burden of researchers.

Enhancing Knowledge Graph Embedding with Hierarchical Self-Attention and Graph Neural Network Techniques for Drug-Drug Interaction Prediction in Virtual Reality Environments

In biomedicine, the critical task is to decode Drug–Drug Interactions (DDIs) from complex biomedical texts. The scientific community employs Knowledge Graph Embedding (KGE) methods, enhanced with advanced neural network technologies, including capsule networks. However, existing methodologies primarily focus on the structural details of individual entities or relations within Biomedical Knowledge Graphs (BioKGs), overlooking the overall structural context of BioKGs, molecular structures, positional features of drug pairs, and their critical Relational Mapping Properties. To tackle the challenges identified, this study presents HSTrHouse an innovative hierarchical self-attention BioKGs embedding framework. This architecture integrates self-attention mechanisms with advanced neural network technologies, including Convolutional Neural Network (CNN) and Graph Neural Network (GNN), for enhanced computational modeling in biomedical contexts. The model bifurcates the BioKGs into entity and relation layers for structural analysis. It employs self-attention across these layers, utilizing PubMedBERT and CNN for position feature extraction, and a GNN for drug pair molecular structure analysis. Then, we connect the position and molecular structure features to integrate them into the self-attention calculation of entity and relation. After that, the output of the self-attention layer is combined with the connected vectors of the position feature and molecular structure feature to obtain the final representation vector, and finally, to model the Relational Mapping Properties (RMPs), the representation vector is embedded into the complex vector space using Householder projections to obtain the BioKGs model. The paper validates HSTrHouse’s efficacy by comparing it with advanced models on three standard BioKGs for DDIs research.

GEHGAN[formula omitted] CircRNA–disease association prediction via graph embedding and heterogeneous graph attention network

There is growing proof suggested that circRNAs play a crucial function in diverse important biological reactions related to human diseases. Within the area of biochemistry, a massive range of wet experiments have been carried out to find out the connections of circRNA–disease in recent years. Since wet experiments are expensive and laborious, nowadays, calculation-based solutions have increasingly attracted the attention of researchers. However, the performance of these methods is restricted due to the inability to balance the distribution among various types of nodes. To remedy the problem, we present a novel computational method called GEHGAN to forecast the new relationships in this research, leveraging graph embedding and heterogeneous graph attention networks. Firstly, we calculate circRNA sequences similarity, circRNA RBP similarity, disease semantic similarity and corresponding GIP kernel similarity to construct heterogeneous graph. Secondly, a graph embedding method using random walks with jump and stay strategies is applied to obtain the preliminary embeddings of circRNAs and diseases, greatly improving the performance of the model. Thirdly, a multi-head graph attention network is employed to further update the embeddings, followed by the employment of the MLP as a predictor. As a result, the five-fold cross-validation indicates that GEHGAN achieves an outstanding AUC score of 0.9829 and an AUPR value of 0.9815 on the CircR2Diseasev2.0 database, and case studies on osteosarcoma, gastric and colorectal neoplasms further confirm the model’s efficacy at identifying circRNA–disease correlations.