Narrow Results

Oil and natural gas are indispensable energy commodities. Investigating the evolution and transformation of energy market dynamics represented by oil and natural gas holds significant implications for global energy governance, as well as political and economic development. We have employed the visibility graph algorithm to transform the most intuitive and readily available time series data of oil and natural gas prices into complex networks. Then, three network embedding algorithms based on machine learning are utilized to embed the network into multidimensional vector spaces, enabling us to identify market state changes in both energy products through the sequential clustering method. We find that the visibility graph algorithm and network embedding methods effectively preserve internal structural characteristics of data, capture similarities between price change rate nodes across different trading days in the oil and gas markets, while also identifying the memorability of price volatility trends. Beyond simple price trends, critical market fluctuations such as those induced by events like the COVID-19 or Russia–Ukraine conflict can be discerned from similarity matrix partitions of network node representation vectors. Furthermore, the sequential clustering algorithm accurately identifies transition points in the oil and gas markets’ states. Unlike conventional time series analysis methods, this innovative combination of visibility graph and network embedding algorithms allows for a multi-faceted exploration of market state changes at various levels; thereby facilitating deeper insights into underlying logic behind price time series.

Graph Edit Distance (GED) computation is a fundamental yet NP-hard problem in graph theory that quantifies the structural dissimilarity between graphs through a series of edit operations. Despite its significance in fields like bioinformatics, cheminformatics, and social network analysis, the computational complexity of exact GED calculation has driven the development of heuristic and approximate methods. This paper proposes a novel approach leveraging Graph Neural Networks to predict GED efficiently. By capturing both local and global graph features through advanced embedding techniques and integrating the Weisfeiler–Lehman graph kernel, our method achieves high accuracy in estimating GED values. Extensive experiments on datasets such as AIDS, Linux, and IMDB demonstrate that our approach outperforms existing methods in terms of mean absolute error (MAE) and computational feasibility. The proposed framework not only enhances the scalability and precision of GED computation, but also provides a robust tool for graph dissimilarity assessments in various application domains.

Urban road network (referred to as the road network) is a complex and highly sparse network. Link prediction of the urban road network can reasonably predict urban structural changes and assist urban designers in decision-making. In this paper, a new link prediction model ASFC is proposed for the characteristics of the road network. The model first performs network embedding on the road network through road2vec algorithm, and then organically combines the subgraph pattern with the network embedding results and the Katz index together, and then we construct the all-order subgraph feature that includes low-order, medium-order and high-order subgraph features and finally to train the logistic regression classification model for road network link prediction. The experiment compares the performance of the ASFC model and other link prediction models in different countries and different types of urban road networks and the influence of changes in model parameters on prediction accuracy. The results show that ASFC performs well in terms of prediction accuracy and stability.

Studies have indicated that focusing solely on pairwise interactions between two nodes disregards the associativity among multi-nodes in the network’s local structure. This associativity can be seen as dependencies among nodes, where certain edges’ presence depends on the path leading to it. Examinations on diverse datasets have approved that the variable order of chained dependencies allows for the preservation of structure information, which enables the reconstruction of the original network into a Higher-Order Network (HON) with improved quality of network representation. This paper proposes a Density-based Higher-Order Network Embedding (DHONE) algorithm, which integrates the concept of higher-order density into the network-embedding process in order to classify the contribution of different orders of dependencies. Through the construction of a novel and effective higher-order adjacency matrix, DHONE steadily improves the accuracy of network representation learning. Experimental results demonstrate DHONEs proficiency in improving embedding accuracy and overall algorithm robustness. Furthermore, grounded in the concept of higher-order density proposed herein, numerous dependencies have been discerned within the network generated from trajectories, potentially indicating the role of multi-node structures in networks.

In network science, link prediction is a technique used to predict missing or future relationships based on currently observed connections. Much attention from the network science community is paid to this direction recently. However, most present approaches predict links based on ad hoc similarity definitions. To address this issue, we propose a link prediction algorithm named Transferring Similarity Based on Adjacency Embedding (TSBAE). TSBAE is based on network embedding, where the potential information of the structure is preserved in the embedded vector space, and the similarity is inherently captured by the distance of these vectors. Furthermore, to accommodate the fact that the similarity should be transferable, indirect similarity between nodes is incorporated to improve the accuracy of prediction. The experimental results on 10 real-world networks show that TSBAE outperforms the baseline algorithms in the task of link prediction, with the cost of tuning a free parameter in the prediction.

Combining Matrix Factorization (MF) with Network Embedding (NE) has been a promising solution to social recommender systems. However, in most of the current combined schemes, the user-specific linking proportions learned by NE are fed to the downstream MF, but not reverse, which is sub-optimal as the rating information is not utilized to discover the linking features for users. Furthermore, the existing combined models mainly focus on enhancing the representation learning for users by exploiting user–user network, yet ignore the representation improvement for items. In this paper, we propose a novel social recommendation scheme, called MF with dual-network collaborative embedding (MF-decoding), which jointly optimizes an integrated objective function of MF and NE, in which both MF and NE tasks can be mutually reinforced in a unified learning process. In particular, the explicit user–user network and the implicit item–item network are collaboratively used by MF-decoding to enhance the representation learning for users and items simultaneously. Our encouraging experimental results on three benchmarks validate the superiority of the proposed MF-decoding model over state-of-the-art social recommendation methods.

Long non-coding RNA (lncRNA), microRNA, and messenger RNA enable key regulations of various biological processes through a variety of diverse interaction mechanisms. Identifying the interactions and cross-talk between these heterogeneous RNA classes is essential in order to uncover the functional role of individual RNA transcripts, especially for unannotated and sparsely discovered RNA sequences with no known interactions. Recently, sequence-based deep learning and network embedding methods are gaining traction as high-performing and flexible approaches that can either predict RNA-RNA interactions from sequence or infer missing interactions from patterns that may exist in the network topology. However, most of the current methods have several limitations, e.g., the inability to perform inductive predictions, to distinguish the directionality of interactions, or to integrate various sequence, interaction, expression, and genomic annotation datasets. We proposed a novel deep learning framework, rna2rna, which learns from RNA sequences to produce a low-dimensional embedding that preserves proximities in both the interaction topology and the functional affinity topology. In this proposed embedding space, the two-part “source and target contexts”capture the receptive fields of each RNA transcript to encapsulate heterogeneous cross-talk interactions between lncRNAs and microRNAs. The proximity between RNAs in this embedding space also uncovers the second-order relationships that allow for accurate inference of novel directed interactions or functional similarities between any two RNA sequences. In a prospective evaluation, our method exhibits superior performance compared to state-of-art approaches at predicting missing interactions from several RNA-RNA interaction databases. Additional results suggest that our proposed framework can capture a manifold for heterogeneous RNA sequences to discover novel functional annotations.

Protein phosphorylation is a key post-translational modification that plays a central role in many cellular processes. With recent advances in biotechnology, thousands of phosphorylated sites can be identified and quantified in a given sample, enabling proteome-wide screening of cellular signaling. However, for most (> 90%) of the phosphorylation sites that are identified in these experiments, the kinase(s) that target these sites are unknown. To broadly utilize available structural, functional, evolutionary, and contextual information in predicting kinase-substrate associations (KSAs), we develop a network-based machine learning framework. Our framework integrates a multitude of data sources to characterize the landscape of functional relationships and associations among phosphosites and kinases. To construct a phosphosite-phosphosite association network, we use sequence similarity, shared biological pathways, co-evolution, co-occurrence, and co-phosphorylation of phosphosites across different biological states. To construct a kinase-kinase association network, we integrate protein-protein interactions, shared biological pathways, and membership in common kinase families. We use node embeddings computed from these heterogeneous networks to train machine learning models for predicting kinase-substrate associations. Our systematic computational experiments using the PhosphositePLUS database shows that the resulting algorithm, NetKSA, outperforms two state-of-the-art algorithms, including KinomeXplorer and LinkPhinder, in overall KSA prediction. By stratifying the ranking of kinases, NetKSA also enables annotation of phosphosites that are targeted by relatively less-studied kinases.

Availability: The code and data are available at compbio.case.edu/NetKSA/.