Pub Date : 2025-07-30DOI: 10.1109/TSIPN.2025.3577317
Hui Li;Lai Wei;Kunquan Li;Guimin Huang;Jun Li
In today’s highly informalized society, the speed and scope of rumor dissemination pose a great threat to social stability and personal interests. Detecting rumors manually requires a lot of human effort. Therefore, automatic rumor detection has received significant attention. Recently, some researchers have focused on using propagation structural features to identify rumors. However, existing propagation structure-based methods either utilize only spatial features or only temporal features of propagation. Few models can effectively leverage both types of propagation structural features. This paper proposes a Source-Guided Temporal-Spatial joint rumor detection model (SGTS). SGTS dynamically divides the propagation process of an event into a series of temporal sub-events. Additionally, SGTS employs an information-level connection strategy that incorporates spatial structural features from previous temporal stages into the encoding of subsequent stages. In this way, SGTS can effectively capture the spatiotemporal features of propagation. Experimental results and in-depth analysis on commonly-used datasets demonstrate that SGTS achieves significant improvements over existing methods.
{"title":"Integrating Temporal and Spatial Structures for Robust Rumor Detection in Social Networks","authors":"Hui Li;Lai Wei;Kunquan Li;Guimin Huang;Jun Li","doi":"10.1109/TSIPN.2025.3577317","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3577317","url":null,"abstract":"In today’s highly informalized society, the speed and scope of rumor dissemination pose a great threat to social stability and personal interests. Detecting rumors manually requires a lot of human effort. Therefore, automatic rumor detection has received significant attention. Recently, some researchers have focused on using propagation structural features to identify rumors. However, existing propagation structure-based methods either utilize only spatial features or only temporal features of propagation. Few models can effectively leverage both types of propagation structural features. This paper proposes a Source-Guided Temporal-Spatial joint rumor detection model (SGTS). SGTS dynamically divides the propagation process of an event into a series of temporal sub-events. Additionally, SGTS employs an information-level connection strategy that incorporates spatial structural features from previous temporal stages into the encoding of subsequent stages. In this way, SGTS can effectively capture the spatiotemporal features of propagation. Experimental results and in-depth analysis on commonly-used datasets demonstrate that SGTS achieves significant improvements over existing methods.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"821-830"},"PeriodicalIF":3.0,"publicationDate":"2025-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144751049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study focuses on addressing distributed Byzantine-resilient output containment issues for heterogeneous continuous-time multi-agent systems. Inspired by the digital twin technology which creates a virtual replica of a physical object or system, a virtual layer named twin layer is introduced in this work, which is parallel to the conventional cyber-physical layer. The twin layer is more secure than the cyber-physical layer, which generates the secure reference trajectory of each agent via real-time data processing and simulation. Moreover, it decouples the resilient output containment against Byzantine attacks (BA) into two defense sub-schemes: One on the twin layer against Byzantine edge attacks (sending wrong and different messages to neighbors) and the other on the cyber-physical layer against Byzantine node attacks (falsifying input signals). On the twin layer, we develop a topology-assignable distributed resilient estimator by utilizing a novel secure centroid approach, which enhances the resilience of the twin layer by adding a minimal fraction of trusted edges. It is proved that achieving strong $[({n+1})f+1]$-robustness towards the leader set is adequate for ensuring the resilience of the twin layer. On the cyber-physical layer, we design a decentralized adaptive controller against Byzantine node attacks and can also handle potential inter-layered controller faults. This novel adaptive controller has the merit of converging exponentially at an adjustable rate, whose error bound can be explicitly stated. Consequently, we manage to address the resilient containment problem against BAs, in which the agents subject to Byzantine node attacks can also achieve output containment instead of just the normal agents. The simulation examples confirm the efficacy of this newly developed hierarchical protocol, where both normal and Byzantine followers converge within the dynamic convex hull formed by the normal leaders.
{"title":"Resilient Output Containment of Heterogeneous Multi-Agent Systems Against Byzantine Attacks","authors":"Xin Gong;Yang Cao;Xiuxian Li;Hong Lin;Zhan Shu;Guanghui Wen","doi":"10.1109/TSIPN.2025.3592314","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3592314","url":null,"abstract":"This study focuses on addressing distributed Byzantine-resilient output containment issues for heterogeneous continuous-time multi-agent systems. Inspired by the digital twin technology which creates a virtual replica of a physical object or system, a virtual layer named twin layer is introduced in this work, which is parallel to the conventional cyber-physical layer. The twin layer is more secure than the cyber-physical layer, which generates the secure reference trajectory of each agent via real-time data processing and simulation. Moreover, it decouples the resilient output containment against Byzantine attacks (BA) into two defense sub-schemes: One on the twin layer against Byzantine edge attacks (sending wrong and different messages to neighbors) and the other on the cyber-physical layer against Byzantine node attacks (falsifying input signals). On the twin layer, we develop a topology-assignable distributed resilient estimator by utilizing a novel secure centroid approach, which enhances the resilience of the twin layer by adding a minimal fraction of trusted edges. It is proved that achieving strong <inline-formula><tex-math>$[({n+1})f+1]$</tex-math></inline-formula>-robustness towards the leader set is adequate for ensuring the resilience of the twin layer. On the cyber-physical layer, we design a decentralized adaptive controller against Byzantine node attacks and can also handle potential inter-layered controller faults. This novel adaptive controller has the merit of converging exponentially at an adjustable rate, whose error bound can be explicitly stated. Consequently, we manage to address the resilient containment problem against BAs, in which the agents subject to Byzantine node attacks can also achieve output containment instead of just the normal agents. The simulation examples confirm the efficacy of this newly developed hierarchical protocol, where both normal and Byzantine followers converge within the dynamic convex hull formed by the normal leaders.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"938-951"},"PeriodicalIF":3.0,"publicationDate":"2025-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144810706","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-24DOI: 10.1109/TSIPN.2025.3592321
Ding Zhang;Jane Downer;Can Chen;Ren Wang
Graph classification is essential for understanding complex biological systems, where molecular structures and interactions are naturally represented as graphs. Traditional graph neural networks (GNNs) perform well on static tasks but struggle in dynamic settings due to catastrophic forgetting. We present Perturbed and Sparsified Continual Graph Learning (PSCGL), a robust and efficient continual graph learning framework for graph data classification, specifically targeting biological datasets. We introduce a perturbed sampling strategy to identify critical data points that contribute to model learning and a motif-based graph sparsification technique to reduce storage needs while maintaining performance. Additionally, our PSCGL framework inherently defends against graph backdoor attacks, which is crucial for applications in sensitive biological contexts. Extensive experiments on biological datasets demonstrate that PSCGL not only retains knowledge across tasks but also enhances the efficiency and robustness of graph classification models in biology.
{"title":"Efficient and Robust Continual Graph Learning for Graph Classification in Biology","authors":"Ding Zhang;Jane Downer;Can Chen;Ren Wang","doi":"10.1109/TSIPN.2025.3592321","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3592321","url":null,"abstract":"Graph classification is essential for understanding complex biological systems, where molecular structures and interactions are naturally represented as graphs. Traditional graph neural networks (GNNs) perform well on static tasks but struggle in dynamic settings due to catastrophic forgetting. We present Perturbed and Sparsified Continual Graph Learning (PSCGL), a robust and efficient continual graph learning framework for graph data classification, specifically targeting biological datasets. We introduce a perturbed sampling strategy to identify critical data points that contribute to model learning and a motif-based graph sparsification technique to reduce storage needs while maintaining performance. Additionally, our PSCGL framework inherently defends against graph backdoor attacks, which is crucial for applications in sensitive biological contexts. Extensive experiments on biological datasets demonstrate that PSCGL not only retains knowledge across tasks but also enhances the efficiency and robustness of graph classification models in biology.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"952-964"},"PeriodicalIF":3.0,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144843091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-24DOI: 10.1109/TSIPN.2025.3592332
Jian Liu;Wenqiang Wang;Bing Hu;Jinliang Liu;Engang Tian;Jie Cao
This article focuses on the probability-constrained distributed non-fragile (PDNF) estimation problem for nonlinear time-varying systems with unknown but bounded noises, sensor saturation and uniform quantization over sensor networks (SNs). Owing to the limited bandwidth resources, stochastic communication protocol (SCP) is employed to manage network transmission and prevent data collision. At each transmission instant, the sensor node is allowed to communicate with only one randomly selected neighboring sensor. Meanwhile, the non-fragility of the estimator is taken into account to handle potential parameter variations. The goal of this article is to develop a PDNF estimation algorithm such that 1) the estimation error is confined within a certain ellipsoidal region with a predefined probability; and 2) the resulting error ellipsoid is minimized in the sense of matrix trace to achieve optimal estimation performance. In light of this, the sufficient criteria for the availability of the estimator are derived through recursive linear matrix inequality (RLMI) technique. Furthermore, the optimal estimator parameters are attained by solving a convex optimization problem. Ultimately, two simulation experiments are presented to validate the feasibility and practicality of the designed estimation algorithm.
{"title":"Probability-Constrained Distributed Non-Fragile Estimation Over Sensor Networks Subject to Stochastic Communication Protocol","authors":"Jian Liu;Wenqiang Wang;Bing Hu;Jinliang Liu;Engang Tian;Jie Cao","doi":"10.1109/TSIPN.2025.3592332","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3592332","url":null,"abstract":"This article focuses on the probability-constrained distributed non-fragile (PDNF) estimation problem for nonlinear time-varying systems with unknown but bounded noises, sensor saturation and uniform quantization over sensor networks (SNs). Owing to the limited bandwidth resources, stochastic communication protocol (SCP) is employed to manage network transmission and prevent data collision. At each transmission instant, the sensor node is allowed to communicate with only one randomly selected neighboring sensor. Meanwhile, the non-fragility of the estimator is taken into account to handle potential parameter variations. The goal of this article is to develop a PDNF estimation algorithm such that 1) the estimation error is confined within a certain ellipsoidal region with a predefined probability; and 2) the resulting error ellipsoid is minimized in the sense of matrix trace to achieve optimal estimation performance. In light of this, the sufficient criteria for the availability of the estimator are derived through recursive linear matrix inequality (RLMI) technique. Furthermore, the optimal estimator parameters are attained by solving a convex optimization problem. Ultimately, two simulation experiments are presented to validate the feasibility and practicality of the designed estimation algorithm.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"888-900"},"PeriodicalIF":3.0,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144781921","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-18DOI: 10.1109/TSIPN.2025.3589685
Guobing Qian;Jiayin Wang;Luping Shen;Ying-Ren Chien;Junhui Qian;Shiyuan Wang
Distributed adaptive filtering has emerged as a critical methodology across diverse application domains, including wireless sensor networks, distributed signal processing, and intelligent control systems. However, existing diffusion-based adaptive filters suffer performance degradation in non-Gaussian noise and complex network topologies, leading to sub-optimal operation and instability risks. These limitations motivate the development of a robust framework that maintains distributed processing advantages while improving noise robustness. To address this, we propose a distributed widely-linear exponential functional link network (D-WLEFLN) combining wide-linear architecture with exponential expansions for enhanced nonlinear modeling. Furthermore, we develop a kernel risk Blake- Zisserman (KRBZ) based cost function to achieve enhanced outlier resilience. Building upon this foundation, a diffusion recursive kernel risk Blake-Zisserman (D-RKRBZ) algorithm is developed through recursive optimization, alongside a computationally efficient variant specifically designed for the WL architecture to maintain operational efficiency while preserving estimation accuracy. We provide theoretical analysis for the proposed algorithm, encompassing both mean stability and mean square performance. Simulation results validate that the performance of the proposed D-RKRBZ algorithm closely aligns with theoretical analysis. Comparative evaluations against existing diffusion counterparts reveal that D-RKRBZ can achieve lower mean square deviation (MSD) in complex-valued non-Gaussian environments, including contaminated Gaussian (CG) noise and $alpha $ stable noise scenarios.
{"title":"Robust Diffusion Recursive Algorithm for Distributed Widely-Linear Exponential Functional Link Network","authors":"Guobing Qian;Jiayin Wang;Luping Shen;Ying-Ren Chien;Junhui Qian;Shiyuan Wang","doi":"10.1109/TSIPN.2025.3589685","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3589685","url":null,"abstract":"Distributed adaptive filtering has emerged as a critical methodology across diverse application domains, including wireless sensor networks, distributed signal processing, and intelligent control systems. However, existing diffusion-based adaptive filters suffer performance degradation in non-Gaussian noise and complex network topologies, leading to sub-optimal operation and instability risks. These limitations motivate the development of a robust framework that maintains distributed processing advantages while improving noise robustness. To address this, we propose a distributed widely-linear exponential functional link network (D-WLEFLN) combining wide-linear architecture with exponential expansions for enhanced nonlinear modeling. Furthermore, we develop a kernel risk Blake- Zisserman (KRBZ) based cost function to achieve enhanced outlier resilience. Building upon this foundation, a diffusion recursive kernel risk Blake-Zisserman (D-RKRBZ) algorithm is developed through recursive optimization, alongside a computationally efficient variant specifically designed for the WL architecture to maintain operational efficiency while preserving estimation accuracy. We provide theoretical analysis for the proposed algorithm, encompassing both mean stability and mean square performance. Simulation results validate that the performance of the proposed D-RKRBZ algorithm closely aligns with theoretical analysis. Comparative evaluations against existing diffusion counterparts reveal that D-RKRBZ can achieve lower mean square deviation (MSD) in complex-valued non-Gaussian environments, including contaminated Gaussian (CG) noise and <inline-formula><tex-math>$alpha $</tex-math></inline-formula> stable noise scenarios.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"845-858"},"PeriodicalIF":3.0,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144750973","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-11DOI: 10.1109/TSIPN.2025.3587400
Feng Ji;Yanan Zhao;See Hian Lee;Kai Zhao;Wee Peng Tay;Jielong Yang
In graph neural networks (GNNs), both node features and labels are examples of graph signals. While it is common in graph signal processing to impose signal smoothness constraints in learning and estimation tasks, it is unclear how this can be done for discrete node labels. We bridge this gap by introducing the concept of graph distributional signals. We work with the distributions of node labels instead of their values and propose notions of smoothness and non-uniformity of such graph distributional signals. We then propose a general regularization method for GNNs that allows us to encode distributional smoothness and non-uniformity of the model output in semi-supervised node classification tasks. Numerical experiments demonstrate that our method can significantly improve the performance of most base GNN models in different problem settings.
{"title":"Graph Distributional Signals for Regularization in Graph Neural Networks","authors":"Feng Ji;Yanan Zhao;See Hian Lee;Kai Zhao;Wee Peng Tay;Jielong Yang","doi":"10.1109/TSIPN.2025.3587400","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3587400","url":null,"abstract":"In graph neural networks (GNNs), both node features and labels are examples of graph signals. While it is common in graph signal processing to impose signal smoothness constraints in learning and estimation tasks, it is unclear how this can be done for discrete node labels. We bridge this gap by introducing the concept of graph distributional signals. We work with the distributions of node labels instead of their values and propose notions of smoothness and non-uniformity of such graph distributional signals. We then propose a general regularization method for GNNs that allows us to encode distributional smoothness and non-uniformity of the model output in semi-supervised node classification tasks. Numerical experiments demonstrate that our method can significantly improve the performance of most base GNN models in different problem settings.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"670-682"},"PeriodicalIF":3.0,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657391","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-10DOI: 10.1109/TSIPN.2025.3587404
Shaojie Yao;Lei Shi;Yong Wang;Rui Liu;Yi Zhou
This paper addresses the problem of cybersecurity in the localization process of mobile sensor networks. A generic model of random periodic denial-of-service (DoS) attacks is considered, where the attacker’s behavior is bounded only by the duration and frequency of the DoS attacks. A class of distributed localization algorithms based on virtual convex hull is proposed, which is abstracted as a linear time-varying system by constructing virtual convex hull using a single anchor node. Using the method combining graph composition and sub-stochastic matrix, it is shown that the algorithm can accurately converge to the true locations of sensor nodes. At last, the effectiveness of the algorithm is verified by simulation examples.
{"title":"Virtual Convex Hull Based Distributed Iterative Localization for Mobile Sensor Networks Under Denial-of-Service Attacks","authors":"Shaojie Yao;Lei Shi;Yong Wang;Rui Liu;Yi Zhou","doi":"10.1109/TSIPN.2025.3587404","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3587404","url":null,"abstract":"This paper addresses the problem of cybersecurity in the localization process of mobile sensor networks. A generic model of random periodic denial-of-service (DoS) attacks is considered, where the attacker’s behavior is bounded only by the duration and frequency of the DoS attacks. A class of distributed localization algorithms based on virtual convex hull is proposed, which is abstracted as a linear time-varying system by constructing virtual convex hull using a single anchor node. Using the method combining graph composition and sub-stochastic matrix, it is shown that the algorithm can accurately converge to the true locations of sensor nodes. At last, the effectiveness of the algorithm is verified by simulation examples.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"780-793"},"PeriodicalIF":3.0,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144687679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-10DOI: 10.1109/TSIPN.2025.3587414
Yaoyao Zhou;Gang Chen;Zhenghua Chen
In the large-scale cyber-physical systems, to conserve communication resources holds critical significance, thereby driving extensive research interest toward distributed optimization algorithms with high communication efficiency. This paper investigates the constrained distributed dynamic parameter estimation problem (CDPE) for communication resource conservation, and further considers how to cope with more generally directed communication structure, unavoidable arbitrary bounded communication delays, and diverse update strategies. We introduce a new process-based triggering strategy and develop an efficient Process-based Triggering Accelerated Dual Averaging Algorithm(PTADA). Compared with the traditional time-dependent threshold, the PTADA can well adapt to the dynamic behavior of distributed optimization and save communication resources. Our dynamic bound is linear and is independent of the explicit time horizon. Moreover, we further extend PTADA to address scenarios where gradient information cannot be directly obtained, while ensuring no performance degradation. This extension can make the algorithm more realistic and universal. Finally, a distributed multi-sensor network is set up to verify the effectiveness of the algorithm.
{"title":"Process-Based Triggering and Accelerated Dual Averaging Algorithm for Dynamic Parameter Estimation","authors":"Yaoyao Zhou;Gang Chen;Zhenghua Chen","doi":"10.1109/TSIPN.2025.3587414","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3587414","url":null,"abstract":"In the large-scale cyber-physical systems, to conserve communication resources holds critical significance, thereby driving extensive research interest toward distributed optimization algorithms with high communication efficiency. This paper investigates the constrained distributed dynamic parameter estimation problem (CDPE) for communication resource conservation, and further considers how to cope with more generally directed communication structure, unavoidable arbitrary bounded communication delays, and diverse update strategies. We introduce a new process-based triggering strategy and develop an efficient Process-based Triggering Accelerated Dual Averaging Algorithm(PTADA). Compared with the traditional time-dependent threshold, the PTADA can well adapt to the dynamic behavior of distributed optimization and save communication resources. Our dynamic bound is linear and is independent of the explicit time horizon. Moreover, we further extend PTADA to address scenarios where gradient information cannot be directly obtained, while ensuring no performance degradation. This extension can make the algorithm more realistic and universal. Finally, a distributed multi-sensor network is set up to verify the effectiveness of the algorithm.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"683-695"},"PeriodicalIF":3.0,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144680837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Identifying protein binding sites, the specific regions on a protein’s surface where interactions with other molecules occur, is crucial for understanding disease mechanisms and facilitating drug discovery. Although numerous computational techniques have been developed to identify protein binding sites, serving as a valuable screening tool that reduces the time and cost associated with conventional experimental approaches, achieving significant improvements in prediction accuracy remains a formidable challenge. Recent advancements in protein structure prediction, notably through tools like AlphaFold, have made vast numbers of 3-D protein structures available, presenting an opportunity to enhance binding site prediction methods. The availability of detailed 3-D structures has led to the development of Equivariant Graph Neural Networks (GNNs), which can analyze complex spatial relationships in protein structures while maintaining invariance to rotations and translations. However, current equivariant GNN methods still face limitations in fully exploiting the geometric features of protein structures. To address this, we introduce E(Q)AGNN-PPIS, an Equivariant Attention-Enhanced Graph Neural Network designed for predicting protein binding sites by leveraging 3-D protein structure. Our method augments the Equivariant GNN framework by integrating an attention mechanism. This attention component allows the model to focus on the most relevant structural features for binding site prediction, significantly enhancing its ability to capture complex spatial patterns and interactions within the protein structure. Our experimental findings underscore the enhanced performance of E(Q)AGNN-PPIS compared to current state-of-the-art approaches, exhibiting gains of 8.33% in the Area Under the Precision-Recall Curve (AUPRC) and 10% in the Matthews Correlation Coefficient (MCC) across benchmark datasets. Additionally, our method demonstrates fast inference and robust generalization across proteins with varying sequence lengths, outperforming baseline methods.
{"title":"E(Q)AGNN-PPIS: Attention Enhanced Equivariant Graph Neural Network for Protein-Protein Interaction Site Prediction","authors":"Animesh;Rishi Suvvada;Plaban Kumar Bhowmick;Pralay Mitra","doi":"10.1109/TSIPN.2025.3587396","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3587396","url":null,"abstract":"Identifying protein binding sites, the specific regions on a protein’s surface where interactions with other molecules occur, is crucial for understanding disease mechanisms and facilitating drug discovery. Although numerous computational techniques have been developed to identify protein binding sites, serving as a valuable screening tool that reduces the time and cost associated with conventional experimental approaches, achieving significant improvements in prediction accuracy remains a formidable challenge. Recent advancements in protein structure prediction, notably through tools like AlphaFold, have made vast numbers of 3-D protein structures available, presenting an opportunity to enhance binding site prediction methods. The availability of detailed 3-D structures has led to the development of Equivariant Graph Neural Networks (GNNs), which can analyze complex spatial relationships in protein structures while maintaining invariance to rotations and translations. However, current equivariant GNN methods still face limitations in fully exploiting the geometric features of protein structures. To address this, we introduce E(Q)AGNN-PPIS, an Equivariant Attention-Enhanced Graph Neural Network designed for predicting protein binding sites by leveraging 3-D protein structure. Our method augments the Equivariant GNN framework by integrating an attention mechanism. This attention component allows the model to focus on the most relevant structural features for binding site prediction, significantly enhancing its ability to capture complex spatial patterns and interactions within the protein structure. Our experimental findings underscore the enhanced performance of E(Q)AGNN-PPIS compared to current state-of-the-art approaches, exhibiting gains of 8.33% in the Area Under the Precision-Recall Curve (AUPRC) and 10% in the Matthews Correlation Coefficient (MCC) across benchmark datasets. Additionally, our method demonstrates fast inference and robust generalization across proteins with varying sequence lengths, outperforming baseline methods.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"740-751"},"PeriodicalIF":3.0,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144687640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-10DOI: 10.1109/TSIPN.2025.3587399
Yacine Mokhtari;Emmanuel Moulay;Patrick Coirault;Jerome Le Ny
Optimal transport on a graph focuses on finding the most efficient way to transfer resources from one distribution to another while considering the graph’s structure. This paper introduces a new distributed algorithm that solves the optimal transport problem on directed, strongly connected graphs, unlike previous approaches which were limited to bipartite graphs. Our algorithm incorporates quadratic regularization and guarantees convergence using the Alternating Direction Method of Multipliers (ADMM). Notably, it proves convergence not only with quadratic regularization but also in cases without it, whereas earlier works required strictly convex objective functions. In this approach, nodes are treated as agents that collaborate through local interactions to optimize the total transportation cost, relying only on information from their neighbors. Through numerical experiments, we show how quadratic regularization affects both convergence behavior and solution sparsity under different graph structures. Additionally, we provide a practical example that highlights the algorithm robustness through its ability to adjust to topological changes in the graph.
{"title":"An ADMM-Based Approach to Quadratically-Regularized Distributed Optimal Transport on Graphs","authors":"Yacine Mokhtari;Emmanuel Moulay;Patrick Coirault;Jerome Le Ny","doi":"10.1109/TSIPN.2025.3587399","DOIUrl":"https://doi.org/10.1109/TSIPN.2025.3587399","url":null,"abstract":"Optimal transport on a graph focuses on finding the most efficient way to transfer resources from one distribution to another while considering the graph’s structure. This paper introduces a new distributed algorithm that solves the optimal transport problem on directed, strongly connected graphs, unlike previous approaches which were limited to bipartite graphs. Our algorithm incorporates quadratic regularization and guarantees convergence using the Alternating Direction Method of Multipliers (ADMM). Notably, it proves convergence not only with quadratic regularization but also in cases without it, whereas earlier works required strictly convex objective functions. In this approach, nodes are treated as agents that collaborate through local interactions to optimize the total transportation cost, relying only on information from their neighbors. Through numerical experiments, we show how quadratic regularization affects both convergence behavior and solution sparsity under different graph structures. Additionally, we provide a practical example that highlights the algorithm robustness through its ability to adjust to topological changes in the graph.","PeriodicalId":56268,"journal":{"name":"IEEE Transactions on Signal and Information Processing over Networks","volume":"11 ","pages":"1005-1014"},"PeriodicalIF":3.0,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144914223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：