Network Topological Features Research Articles

Optimizing Ecological Spatial Network Topology for Enhanced Carbon Sequestration in the Ecologically Sensitive Middle Reaches of the Yellow River, China

The destruction of vegetation structure and quantity leads to the weakening of the carbon sequestration capacity of the ecosystem. Building an ecological spatial network is a potent method for studying vegetation spatial distribution structures. The relationship between the spatial distribution structure of vegetation networks and carbon sequestration, as approached from the perspective of complex network theory, is understudied. This study uses the minimum resistance model (MCR) and morphological spatial pattern analysis (MSPA) to study the eco-space network and ecological node spatial structure and topological characteristics of vegetation in the ecologically sensitive area of the middle reaches of the Yellow River (ESAMRYR). Based on the Carnegie-Ames-Stanford approach (CASA) model, the vegetation Net Primary Productivity (NPP) of the study area is calculated, and the ecological carbon sequestration function of the ecological node is estimated, and the relationship between the ecological node and the topological indicators is analyzed. The study shows that the forest land carbon storage in the regions situated toward the south and east of the Yellow River ecologically sensitive area is the highest, accounting for twice the proportion of the area, and is very important in terms of increasing carbon storage. Most of the ecological sources in the study area have a higher topological importance than functional importance, and the sources with low coordination are mainly distributed in the southwest and northeast. We construct a topology and function coupling optimization model (TFCO) to explore the coordination between vegetation structure and carbon sequestration function, to determine the network optimization direction, and to propose optimization solutions. Analysis of network robustness and carbon sequestration capacity shows that the sturdiness and carbon sequestration of the enhanced network are significantly improved. This study provides strategies and methods for protecting ecological sensitive areas, optimizing vegetation spatial distribution, and enhancing carbon sequestration capacity.

EBD: an eye biomarker database.

Many ophthalmic disease biomarkers have been identified through comprehensive multiomics profiling, and hold significant potential in advancing the diagnosis, prognosis, and management of diseases. Meanwhile, the eye itself serves as a natural biomarker for several systemic diseases including neurological, renal, and cardiovascular systems. We aimed to collect and standardize this eye biomarkers information and construct the eye biomarker database (EBD) to provide ophthalmologists with a platform to search, analyze, and download these eye biomarker data. In this study, we present the EBD <http://www.eyeseeworld.com/ebd/index.html>, a world-first online compilation comprising 889 biomarkers for 26 ocular diseases and 939 eye biomarkers for 181 systemic diseases. The EBD also includes the information of 78 "nonbiomarkers"-the objects that have been proven cannot be biomarkers. Biological function and network analysis were conducted for these ocular disease biomarkers, and several hub pathways and common network topology characteristics were newly identified, which may promote future ocular disease biomarker discovery and characterizes the landscape of biomarkers for eye diseases at the pathway and network level. The EBD is expected to yield broader utility among developmental biologists and clinical scientists in and outside of the eye field by assisting in the identification of biomarkers linked to eye disorders and related systemic diseases. EBD is available at http://www.eyeseeworld.com/ebd/index.html.

Altered Resting-State Functional Networks in Nondialysis Patients with Stage 5 Chronic Kidney Disease: A Graph–Theoretical Analysis

This study aimed to investigate the topological characteristics of the resting-state functional network and the underlying pathological mechanism in nondialysis patients with stage 5 chronic kidney disease (CKD5 ND). Eighty-five subjects (21 patients with CKD5 ND, 32 patients with CKD on maintenance hemodialysis (HD), and 32 healthy controls (HCs)) underwent laboratory examinations, neuropsychological tests, and brain magnetic resonance imaging. The topological characteristics of networks were compared with a graph-theoretical approach, and correlations between neuropsychological scores and network properties were analyzed. All participants exhibited networks with small-world attributes, and global topological attributes were impaired in both groups of patients with CKD 5 (ND and HD) compared with HCs (p < 0.05); these impairments were more severe in the CKD5 ND group than in the HD group (p < 0.05). Compared with the HC group, the degree centrality of the CKD5 ND group decreased mainly in the basal ganglia and increased in the bilateral orbitofrontal gyrus, bilateral precuneus, and right cuneus. Correlation analysis showed that the degree of small-worldness, normalized clustering coefficients, and Montreal Cognitive Assessment (MoCA) scores were positively correlated and that characteristic path length was negatively correlated with these variables in patients with CKD5 ND. The nodal efficiency of the bilateral putamen (r = 0.53, p < 0.001 and r = 0.47, p < 0.001), left thalamus (r = 0.37, p < 0.001), and right caudate nucleus (r = 0.28, p = 0.01) was positively correlated with MoCA scores. In conclusion, all CKD5 ND patients exhibited changes in functional network topological properties and were closely associated with mild cognitive impairment. More interestingly, the topological property changes in CKD5 ND patients were dominated by basal ganglia areas, which may be more helpful to understand and possibly reveal the underlying pathological mechanisms of cognitive impairment in CKD5 ND.

Attention‐based network embedding with higher‐order weights and node attributes

AbstractNetwork embedding aspires to learn a low‐dimensional vector of each node in networks, which can apply to diverse data mining tasks. In real‐life, many networks include rich attributes and temporal information. However, most existing embedding approaches ignore either temporal information or network attributes. A self‐attention based architecture using higher‐order weights and node attributes for both static and temporal attributed network embedding is presented in this article. A random walk sampling algorithm based on higher‐order weights and node attributes to capture network topological features is presented. For static attributed networks, the algorithm incorporates first‐order to k‐order weights, and node attribute similarities into one weighted graph to preserve topological features of networks. For temporal attribute networks, the algorithm incorporates previous snapshots of networks containing first‐order to k‐order weights, and nodes attribute similarities into one weighted graph. In addition, the algorithm utilises a damping factor to ensure that the more recent snapshots allocate a greater weight. Attribute features are then incorporated into topological features. Next, the authors adopt the most advanced architecture, Self‐Attention Networks, to learn node representations. Experimental results on node classification of static attributed networks and link prediction of temporal attributed networks reveal that our proposed approach is competitive against diverse state‐of‐the‐art baseline approaches.

Node Essentiality Assessment and Distributed Collaborative Virtual Network Embedding in Datacenters

Network virtualization (NV) has extensive and significant applications in cloud computing and parallel and distributed systems. Virtual network embedding (VNE) is a key issue in NV, which is an effective means to advance systems' performance. While existing VNE research lacks resource allocation coordination between mappings of different virtual network requests, resulting in insufficient resource utilization and high overhead. In this paper, we propose a novel node essentiality evaluation model for data center networks (DCNs), and design an efficient distributed collaborative virtual network embedding. Firstly, we propose a node essentiality evaluation scheme based on dynamic model, which combines the characteristics of network topology and nodes to make the evaluation results more comprehensive. Secondly, we establish the two-stage node importance evaluation criteria for the deviation mean of the data center dynamic model and the variance based on the deviation mean. Furthermore, we investigate a nodal importance assessment method based on the data center dynamic model for perturbation testing. Finally, we design a distributed coordinated VNE algorithm ( <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CNI-VNE</b> ) which calculates the importance index of physical nodes through topology awareness. The proposed algorithm can increase the coordination between different request mappings, thereby reducing the mapping cost of physical node resources and minimizing the cost of VNE. We use the real Fat-tree DCN of 128 servers and 80 switches as testbed, and evaluate them from indicators such as average reliability, average bandwidth consumption, average energy consumption, and average mapping time. Massive simulation results in different scenarios show that our algorithm achieves the best performance on most indicators compared with the existing state-of-the-art proposals, mapping acceptance and average revenue increased by 19.4% and 21.3%, respectively, and DCN reduced bandwidth consumption by about 30%.

Brain working memory network indices as landmarks of intelligence

Identifying the neural correlates of intelligence has long been a goal in neuroscience. Recently, the field of network neuroscience has attracted researchers' attention as a means for answering this question. In network neuroscience, the brain is considered as an integrated system whose systematic properties provide profound insights into health and behavioral outcomes. However, most network studies of intelligence have used univariate methods to investigate topological network measures, with their focus limited to a few measures. Furthermore, most studies have focused on resting state networks despite the fact that brain activation during working memory tasks has been linked to intelligence. Finally, the literature is still missing an investigation of the association between network assortativity and intelligence. To address these issues, here we employ a recently developed mixed-modeling framework for analyzing multi-task brain networks to elucidate the most critical working memory task network topological properties corresponding to individuals' intelligence differences. We used a data set of 379 subjects (22–35 y/o) from the Human Connectome Project (HCP). Each subject's data included composite intelligence scores, and fMRI during resting state and a 2-back working memory task. Following comprehensive quality control and preprocessing of the minimally preprocessed fMRI data, we extracted a set of the main topological network features, including global efficiency, degree, leverage centrality, modularity, and clustering coefficient. The estimated network features and subject's confounders were then incorporated into the multi-task mixed-modeling framework to investigate how brain network changes between working memory and resting state relate to intelligence score. Our results indicate that the general intelligence score (cognitive composite score) is associated with a change in the relationship between connection strength and multiple network topological properties, including global efficiency, leverage centrality, and degree difference during working memory as it is compared to resting state. More specifically, we observed a higher increase in the positive association between global efficiency and connection strength for the high intelligence group when they switch from resting state to working memory. The strong connections might form superhighways for a more efficient global flow of information through the brain network. Furthermore, we found an increase in the negative association between degree difference and leverage centrality with connection strength during working memory tasks for the high intelligence group. These indicate higher network resilience and assortativity along with higher circuit-specific information flow during working memory for those with a higher intelligence score. Although the exact neurobiological implications of our results are speculative at this point, our results provide evidence for the significant association of intelligence with hallmark properties of brain networks during working memory.

CRIME HOTSPOT EMERGENCE IN MEXICO CITY: A COMPLEXITY SCIENCE PERSPECTIVE

We present a dynamic model based on the theories proposed by environmental criminologists to explain the emergence of crime hotspots within cities; a pervasive phenomenon that is largely independent of cities size and cultural differences. The model is defined on a multiplex network that represents a city spatial tiling with its corresponding urban transport infrastructure, allowing to explore the relation between crime hotspot locations and the network topological features. It also allows to explore the effects that cities time evolution and police checkpoints might have on the emergence of crime hotspots. For Mexico City, the model shows that heterogeneous distributions of criminal activity arise from a diffusion-driven instability, as a self-organizing process. The results obtained for this city are in line with several insights from environmental criminology, such as the relationship between urban layout and crime hotspots locations, or the conceptual label assigned to specific locations as crime generators. They also uncover new relationships between cities design and crime hotspot locations, and suggest that routine activity theory alone cannot explain the emergence of heterogeneous crime distributions.

Traffic-aware service relocation in software-defined and intent-based elastic optical networks

The paper focuses on the efficient dynamic routing of unicast and data center (dc)-related requests in elastic optical networks (eons) implementing software-defined networking (sdn) and intend-based networking (ibn) paradigms. To improve the network performance (measured as a ratio of the accepted traffic), we apply the service relocation (i.e., the adaptive process of changing the assigned dc for an anycast request). To enable a realistic case study, we propose a novel traffic model for dc-oriented and intent-based transport networks. The model reflects patterns observed in real networks and relates them with the economic and demographic parameters of the cities associated with network nodes. Then, we propose a dedicated allocation algorithm and introduce 21 different service relocation policies. These are traffic- and network-aware approaches, which use three data types for decision-making — traffic prediction, bit-rate rejection history, and topological network characteristics. Finally, we perform extensive simulations to: (i) tune the proposed optimization approaches, (ii) compare their efficiency and select the best one, (iii) determine benefits provided by the traffic- and network-aware service relocation in sdn/ibn optical networks. The results prove a high efficiency of the proposed policies, which allowed to serve up to 5.39% more traffic compared to the network with the fixed dc assignment. They also reveal that the most efficient policy makes its decisions based on traffic prediction.

Motor imagery classification method based on relative wavelet packet entropy brain network and improved lasso.

Motor imagery (MI) electroencephalogram (EEG) signals have a low signal-to-noise ratio, which brings challenges in feature extraction and feature selection with high classification accuracy. In this study, we proposed an approach that combined an improved lasso with relief-f to extract the wavelet packet entropy features and the topological features of the brain function network. For signal denoising and channel filtering, raw MI EEG was filtered based on an R2 map, and then the wavelet soft threshold and one-to-one multi-class score common spatial pattern algorithms were used. Subsequently, the relative wavelet packet entropy and corresponding topological features of the brain network were extracted. After feature fusion, mutcorLasso and the relief-f method were applied for feature selection, followed by three classifiers and an ensemble classifier, respectively. The experiments were conducted on two public EEG datasets (BCI Competition III dataset IIIa and BCI Competition IV dataset IIa) to verify this proposed method. The results showed that the brain network topology features and feature selection methods can retain the information of EEG more effectively and reduce the computational complexity, and the average classification accuracy for both public datasets was above 90%; hence, this algorithms is suitable in MI-BCI and has potential applications in rehabilitation and other fields.

A novel algorithm for identifying influential nodes in complex networks based on local propagation probability model

Identifying influential nodes in complex networks is an essential research issue in network science since it may attribute to understand network structure and function. Majority of existing methods have been established by considering topological characteristics of networks. In this paper, we put forward a novel algorithm based on the Local Propagation Probability (LPP) model for identifying influential nodes in complex networks. The core idea of LPP algorithm is that the nodal influence is measured by total comprehensive scores of neighbor nodes within its three level neighborhood. Specially, the comprehensive score is calculated from three dimensions, namely the propagation influence score between different order neighbors, the propagation influence score in the same order neighbors and the hierarchical structure information of nodes. To validate the performance and applicability of the proposed algorithm, LPP is compared with eight state-of-the-art and competitive algorithms on nine real-world networks. Experimental results demonstrate that LPP performs better in terms of ranking accuracy, effectiveness, top-k nodes and distinguishing ability. The low time complexity allows LPP to be applied to large-scale sparse networks.

Relating Spatial Quality of Public Transportation and the Most Visited Museums: Revisiting Sustainable Mobility of Waterfronts and Historic Centers in International Cruise Destinations

Museums are main tourist resources for independent cruise passengers in cultural cruise destinations. However, their influence on cruise destinations is scarcely analyzed. The aim of this paper was to focus on two questions: What distribution of museums facilitates a sustainable mobility of cruise tourists in balance with urban needs? And which factors affect the potential use of both means of sustainable travel—soft mobility and public transport—at cruise destinations? The network topological features of main museums were analyzed to compare similar features among tourism destinations. A topological study based on the model of the three urban fabrics was related to the objective quality of PT. This allowed for a greater walkability or potential use of PT to be estimated. The results allowed for a diverse sample of cultural tourism port cities with cruise activity to be classified in three centralized levels and two decentralized ones. Results indicated that centralized networks are more prone to cruise tourists, while decentralized networks are more related to main cultural destinations. Finally, the discussion section analysed recommendations and measures to improve sustainable mobility and the planning of new museums. The results of this paper will be of interest to cultural and transport managers at these types of destinations.

Study of brain network alternations in non-lesional epilepsy patients by BOLD-fMRI.

To investigate the changes of brain network in epilepsy patients without intracranial lesions under resting conditions. Twenty-six non-lesional epileptic patients and 42 normal controls were enrolled for BOLD-fMRI examination. The differences in brain network topological characteristics and functional network connectivity between the epilepsy group and the healthy controls were compared using graph theory analysis and independent component analysis. The area under the curve for local efficiency was significantly lower in the epilepsy patients compared with healthy controls, while there were no differences in global indicators. Patients with epilepsy had higher functional connectivity in 4 connected components than healthy controls (orbital superior frontal gyrus and medial superior frontal gyrus, medial superior frontal gyrus and angular gyrus, superior parietal gyrus and paracentral lobule, lingual gyrus, and thalamus). In addition, functional connectivity was enhanced in the default mode network, frontoparietal network, dorsal attention network, sensorimotor network, and auditory network in the epilepsy group. The topological characteristics and functional connectivity of brain networks are changed in in non-lesional epilepsy patients. Abnormal functional connectivity may suggest reduced brain efficiency in epilepsy patients and also may be a compensatory response to brain function early at earlier stages of the disease.

Power Grid Structure Performance Evaluation Based on Complex Network Cascade Failure Analysis

A safe and stable operation power system is very important for the maintenance of national industrial security and social economy. However, with the increasing complexity of the power grid topology and its operation, new challenges in estimating and evaluating the grid structure performance have received significant attention. Complex network theory transfers the power grid to a network with nodes and links, which helps evaluate the system conveniently with a global view. In this paper, we employ the complex network method to address the cascade failure process and grid structure performance assessment simultaneously. Firstly, a grid cascade failure model based on network topology and power system characteristics is constructed. Then, a set of performance evaluation indicators, including invulnerability, reliability, and vulnerability, is proposed based on the actual functional properties of the grid by renewing the power-weighted degree, medium, and clustering coefficients according to the network cascade failure. Finally, a comprehensive network performance evaluation index, which combines the invulnerability, reliability, and vulnerability indicators with an entropy-based objective weighting method, is put forward in this study. In order to confirm the approach’s efficacy, an IEEE-30 bus system is employed for a case study. Numerical results show that the weighted integrated index with a functional network could better evaluate the power grid performance than the unweighted index with a topology network, which demonstrates and validates the effectiveness of the method proposed in this paper.

Associations between psychotropic drugs and rsEEG connectivity and network characteristics: a cross-sectional study in hospital-admitted psychiatric patients.

Resting-state EEG (rsEEG) characteristics, such as functional connectivity and network topology, are studied as potential biomarkers in psychiatric research. However, the presence of psychopharmacological treatment in study participants poses a potential confounding factor in biomarker research. To address this concern, our study aims to explore the impact of both single and multi-class psychotropic treatments on aforementioned rsEEG characteristics in a psychiatric population. RsEEG was analyzed in a real-world cross-sectional sample of 900 hospital-admitted psychiatric patients. Patients were clustered into eight psychopharmacological groups: unmedicated, single-class treatment with antipsychotics (AP), antidepressants (AD) or benzodiazepines (BDZ), and multi-class combinations of these treatments. To assess the associations between psychotropic treatments and the macroscale rsEEG characteristics mentioned above, we employed a general linear model with post-hoc tests. Additionally, Spearman's rank correlation analyses were performed to explore potential dosage effects. Compared to unmedicated patients, single-class use of AD was associated with lower functional connectivity in the delta band, while AP was associated with lower functional connectivity in both the delta and alpha bands. Single-class use of BDZ was associated with widespread rsEEG differences, including lower functional connectivity across frequency bands and a different network topology within the beta band relative to unmedicated patients. All of the multi-class groups showed associations with functional connectivity or topology measures, but effects were most pronounced for concomitant use of all three classes of psychotropics. Differences were not only observed in comparison with unmedicated patients, but were also evident in comparisons between single-class, multi-class, and single/multi-class groups. Importantly, multi-class associations with rsEEG characteristics were found even in the absence of single-class associations, suggesting potential cumulative or interaction effects of different classes of psychotropics. Dosage correlations were only found for antipsychotics. Our exploratory, cross-sectional study suggests small but significant associations between single and multi-class use of antidepressants, antipsychotics and benzodiazepines and macroscale rsEEG functional connectivity and network topology characteristics. These findings highlight the importance of considering the effects of specific psychotropics, as well as their interactions, when investigating rsEEG biomarkers in a medicated psychiatric population.

A Geographic Routing Protocol Based on Node Scoring and Intersection Evaluation

VANET has a good application prospect in the current vehicle communication. As one of MANET, it has the characteristics of dynamic network topology and indirect connection, which makes the wireless connection between nodes unstable. Aiming at this problem, we propose a routing protocol combining link stability, node density and distance scale. When data packets are transmitted to the intersection, the optimal adjacent intersection is selected to determine the forwarding path, which improves the geographical forwarding of data packets located at the intersection. Through simulation, we find that the proposed geographic routing protocol based on node scoring and intersection evaluation (GRNSIE) has good routing performance in PDR and latency when compared with GPSR, GyTAR and GROOV.

Decreased precipitation reduced the complexity and stability of bacterial co-occurrence patterns in a semiarid grassland.

Grasslands harbor complex bacterial communities, whose dynamic interactions are considered critical for organic matter and nutrient cycling. However, less is known about how changes in precipitation impact bacterial interactions. We conducted precipitation manipulation experiments in the Eastern Eurasian Steppe in China and constructed co-occurrence networks for bacterial communities. The network topological features of the bacterial communities exhibited considerable differences among increased precipitation, control, and decreased precipitation gradients. The bacterial co-occurrence pattern in the increased precipitation gradient was the most complex and stable, with a large network size, followed by those of the control and decreased precipitation gradients. Soil moisture (SM) was the primary factor influencing the complexity, size, and stability of bacterial networks across different precipitation gradients, followed by total nitrogen (TN), belowground biomass, aboveground biomass, and total carbon (TC). Our results indicate that drought conditions reduce the complexity and stability of the bacterial community, and future changes in precipitation will greatly reshape bacterial interactions in semiarid grasslands. Overall, these findings could enhance our understanding of how microbes respond to changing precipitation patterns by regulating their interactions in water-limited ecosystems and will improve our ability to predict the impacts of precipitation regime change on ecosystem nutrient cycling and feedback between ecosystem processes and global climate change.

Topological network features determine convergence rate of distributed average algorithms

Gossip algorithms are message-passing schemes designed to compute averages and other global functions over networks through asynchronous and randomised pairwise interactions. Gossip-based protocols have drawn much attention for achieving robust and fault-tolerant communication while maintaining simplicity and scalability. However, the frequent propagation of redundant information makes them inefficient and resource-intensive. Most previous works have been devoted to deriving performance bounds and developing faster algorithms tailored to specific structures. In contrast, this study focuses on characterising the effect of topological network features on performance so that faster convergence can be engineered by acting on the underlying network rather than the gossip algorithm. The numerical experiments identify the topological limiting factors, the most predictive graph metrics, and the most efficient algorithms for each graph family and for all graphs, providing guidelines for designing and maintaining resource-efficient networks. Regression analyses confirm the explanatory power of structural features and demonstrate the validity of the topological approach in performance estimation. Finally, the high predictive capabilities of local metrics and the possibility of computing them in a distributed manner and at a low computational cost inform the design and implementation of a novel distributed approach for predicting performance from the network topology.

Unveiling Qzone: A measurement study of a large-scale online social network

Nowadays, Online Social Network (OSN) is very important in maintaining relationships and disseminating information. Qzone, as one of the most popular OSNs in China, lacks detailed quantitative measurement and analysis. To the best of our knowledge, we are the first to conduct a large-scale measurement on Qzone. Firstly, we develop a dedicated Qzone Crawler System (Qzone-CS) and collect over a billion user and topology data. Subsequently, we measure user behavior and network topology characteristics by analyzing their posting and interactive behaviors. The results show that: 1) the temporal distribution of user behavior follows the periodic distribution, while user posting behavior follows the segmented power-law distribution. Among all the user behaviors on Qzone, the probability of posting Shuoshuo is the highest. Meanwhile, users prefer to comment on others’ blogs than to forward them. Unlike other OSNs, Qzone is a relatively closed system, which is not conducive to the dissemination of information. 2) Qzone is a small-world and scale-free community; the friendship network has higher clustering characteristic compared with the interactive network; out-degree in the interactive network has an obvious correlation with in-degree; and unlike other OSNs, Qzone is a heterogeneous network. Our results can help researchers better understand the user behavior and network topology characteristics of Qzone.

Temporal dynamics of soil bacterial network regulate soil resistomes.

Soil bacteria are diverse and form complicated ecological networks through various microbial interactions, which play important roles in soil multi-functionality. However, the seasonal effects on the bacterial network, especially the relationship between bacterial network topological features and soil resistomes remains underexplored, which impedes our ability to unveil the mechanisms of the temporal-dynamics of antibiotic resistance genes (ARGs). Here, a field investigation was conducted across four seasons at the watershed scale. We observed significant seasonal variation in bacterial networks, with lower complexity and stability in autumn, and a wider bacterial community niche in summer. Similar to bacterial communities, the co-occurrence networks among ARGs also shift with seasonal change, particularly with respect to the topological features of the node degree, which on average was higher in summer than in the other seasons. Furthermore, the nodes with higher betweenness, stress, degree, and closeness centrality in the bacterial network showed strong relationships with the 10 major classes of ARGs. These findings highlighted the changes in the topological properties of bacterial networks that could further alter antibiotic resistance in soil. Together, our results reveal the temporal dynamics of bacterial ecological networks at the watershed scale, and provide new insights into antibiotic resistance management under environmental changes.

Drug response prediction using graph representation learning and Laplacian feature selection

Knowing the responses of a patient to drugs is essential to make personalized medicine practical. Since the current clinical drug response experiments are time-consuming and expensive, utilizing human genomic information and drug molecular characteristics to predict drug responses is of urgent importance. Although a variety of computational drug response prediction methods have been proposed, their effectiveness is still not satisfying. In this study, we propose a method called LGRDRP (Learning Graph Representation for Drug Response Prediction) to predict cell line-drug responses. At first, LGRDRP constructs a heterogeneous network integrating multiple kinds of information: cell line miRNA expression profiles, drug chemical structure similarity, gene-gene interaction, cell line-gene interaction and known cell line-drug responses. Then, for each cell line, learning graph representation and Laplacian feature selection are combined to obtain network topology features related to the cell line. The learning graph representation method learns network topology structure features, and the Laplacian feature selection method further selects out some most important ones from them. Finally, LGRDRP trains an SVM model to predict drug responses based on the selected features of the known cell line-drug responses. Our five-fold cross-validation results show that LGRDRP is significantly superior to the art-of-the-state methods in the measures of the average area under the receiver operating characteristics curve, the average area under the precision-recall curve and the recall rate of top-k predicted sensitive cell lines. Our results demonstrated that the usage of multiple types of information about cell lines and drugs, the learning graph representation method, and the Laplacian feature selection is useful to the improvement of performance in predicting drug responses. We believe that such an approach would be easily extended to similar problems such as miRNA-disease relationship inference.