Graph Convolutional Network Model Research Articles

Learning Aligned Vertex Convolutional Networks for Graph Classification.

Graph convolutional networks (GCNs) are powerful tools for graph structure data analysis. One main drawback arising in most existing GCN models is that of the oversmoothing problem, i.e., the vertex features abstracted from the existing graph convolution operation have previously tended to be indistinguishable if the GCN model has many convolutional layers (e.g., more than two layers). To address this problem, in this article, we propose a family of aligned vertex convolutional network (AVCN) models that focus on learning multiscale features from local-level vertices for graph classification. This is done by adopting a transitive vertex alignment algorithm to transform arbitrary-sized graphs into fixed-size grid structures. Furthermore, we define a new aligned vertex convolution operation that can effectively learn multiscale vertex characteristics by gradually aggregating local-level neighboring aligned vertices residing on the original grid structures into a new packed aligned vertex. With the new vertex convolution operation to hand, we propose two architectures for the AVCN models to extract different hierarchical multiscale vertex feature representations for graph classification. We show that the proposed models can avoid iteratively propagating redundant information between specific neighboring vertices, restricting the notorious oversmoothing problem arising in most spatial-based GCN models. Experimental evaluations on benchmark datasets demonstrate the effectiveness.

ScaleGCN: Efficient and Effective Graph Convolution via Channel-Wise Scale Transformation.

Graph convolutional networks (GCNs) have shown success in many graph-based applications as they can combine node features and graph topology to obtain expressive embeddings. While there exist numerous GCN variants, a typical graph convolution layer uses neighborhood aggregation and fully-connected (FC) layers to extract topological and node-wise features, respectively. However, when the receptive field of GCNs becomes larger, the tight coupling between the number of neighborhood aggregation and FC layers can increase the risk of over-fitting. Also, the FC layer between two successive aggregation operations will mix and pollute features in different channels, bringing noise and making node features hard to converge at each channel. In this article, we explore graph convolution without FC layers. We propose scale graph convolution, a new graph convolution using channel-wise scale transformation to extract node features. We provide empirical evidence that our new method has lower over-fitting risk and needs fewer layers to converge. We show from both theoretical and empirical perspectives that models with scale graph convolution have lower computational and memory costs than traditional GCN models. Experimental results on various datasets show that our method can achieve state-of-the-art results, in a cost-effective fashion.

Clinical prediction of microvascular invasion in hepatocellular carcinoma using an MRI-based graph convolutional network model integrated with nomogram.

Based on enhanced MRI, a prediction model of microvascular invasion (MVI) for hepatocellular carcinoma (HCC) was developed using graph convolutional network (GCN) combined nomogram. We retrospectively collected 182 HCC patients confirmed histopathologically, all of them performed enhanced MRI before surgery. The patients were randomly divided into training and validation groups. Radiomics features were extracted from the arterial phase (AP), portal venous phase (PVP), and delayed phase (DP), respectively. After removing redundant features, the graph structure by constructing the distance matrix with the feature matrix was built. Screening the superior phases and acquired GCN Score (GS). Finally, combining clinical, radiological and GS established the predicting nomogram. 27.5% (50/182) patients were with MVI positive. In radiological analysis, intratumoural artery (P = 0.007) was an independent predictor of MVI. GCN model with grey-level cooccurrence matrix-grey-level run length matrix features exhibited area under the curves of the training group was 0.532, 0.690, and 0.885 and the validation group was 0.583, 0.580, and 0.854 for AP, PVP, and DP, respectively. DP was selected to develop final model and got GS. Combining GS with diameter, corona enhancement, mosaic architecture, and intratumoural artery constructed a nomogram which showed a C-index of 0.884 (95% CI: 0.829-0.927). The GCN model based on DP has a high predictive ability. A nomogram combining GS, clinical and radiological characteristics can be a simple and effective guiding tool for selecting HCC treatment options. GCN based on MRI could predict MVI on HCC.

Supervised contrastive learning enhances graph convolutional networks for predicting neurodevelopmental deficits in very preterm infants using brain structural connectome

Very preterm (VPT) infants (born at less than 32 weeks gestational age) are at high risk for various adverse neurodevelopmental deficits. Unfortunately, most of these deficits cannot be accurately diagnosed until the age of 2-5 years old. Given the benefits of early interventions, accurate diagnosis and prediction soon after birth are urgently needed for VPT infants. Previous studies have applied deep learning models to learn the brain structural connectome (SC) to predict neurodevelopmental deficits in the preterm population. However, none of these models are specifically designed for graph-structured data, and thus may potentially miss certain topological information conveyed in the brain SC. In this study, we aim to develop deep learning models to learn the SC acquired at term-equivalent age for early prediction of neurodevelopmental deficits at 2 years corrected age in VPT infants. We directly treated the brain SC as a graph, and applied graph convolutional network (GCN) models to capture complex topological information of the SC. In addition, we applied the supervised contrastive learning (SCL) technique to mitigate the effects of the data scarcity problem, and enable robust training of GCN models. We hypothesize that SCL will enhance GCN models for early prediction of neurodevelopmental deficits in VPT infants using the SC. We used a regional prospective cohort of ∼280 VPT infants who underwent MRI examinations at term-equivalent age from the Cincinnati Infant Neurodevelopment Early Prediction Study (CINEPS). These VPT infants completed neurodevelopmental assessment at 2 years corrected age to evaluate cognition, language, and motor skills. Using the SCL technique, the GCN model achieved mean areas under the receiver operating characteristic curve (AUCs) in the range of 0.72∼0.75 for predicting three neurodevelopmental deficits, outperforming several competing models. Our results support our hypothesis that the SCL technique is able to enhance the GCN model in our prediction tasks.

Abstract 4945: Computational modeling-based discovery of novel anticancer drug candidates targeting centromere-associated protein E

Abstract Purpose: Centromere-associated protein E (CENP-E), a mitotic motor protein, is crucial for cell division. Inhibition of CENP-E can lead to chromosome misalignment and aneuploidy. Our recent studies revealed that the CENP-E inhibitor (CENP-Ei), Compound A, induces aneuploid-mediated cGAS-STING pathway activation in cancer cell lines, which is expected to trigger innate immune response and induce immunological conversion of the tumor microenvironment from cold to hot. However, known CENP-Eis exhibited limited efficacy, and none have gone beyond phase I trials since 2012. The aim of this study is to develop a workflow integrating structure and artificial intelligence-based modeling approaches to accelerate the discovery of novel CENP-Eis with improved efficacy and overcome related challenges. Method: The inhibitor-bound CENP-E structure was modeled using chimeric homology modeling followed by induced fit docking with known CENP-Eis. Structure-based virtual screening (VS) was performed on large compound libraries in Enamine REAL, Synthetically Accessible Virtual Inventory, and MolPort databases. Consensus scoring methods were applied by averaging ranks of individual molecules obtained from VS against different CENP-E conformations. An active learning-enhanced VS pipeline was developed using ATOM Modeling PipeLine (AMPL) to find the best-ranked ligands more efficiently. CENP-E ligand candidates were further filtered based on their safety properties predicted by machine learning models trained with curated databases of adverse drug reactions using AMPL. The success of this study was measured by its ability to reveal novel chemotypes that could modulate CENP-E in kinesin ATPase assays. Result: Conformations of CENP-E’s ligand binding site were predicted for VS using the homology model. Structure-based VS was performed to select high-ranked compounds based on consensus scoring. The active learning-enhanced VS pipeline achieved 75% accuracy in ranking the top 20% of hit compounds. The graph convolutional network model with a ROCAUC, 0.76, was created using AMPL to filter ligand candidates based on their predicted safety properties. Seventy-three potential CENP-E ligands computationally prioritized from the Molport database were experimentally tested using kinesin ATPase assays from which 4 potential inhibitors and 8 potential activators were identified indicating overlapping binding sites of inhibitors and activators. Conclusion: Our integrated workflow is effective in discovering novel CENP-E modulators. The discovery of CENP-E activators provides an opportunity to identify novel therapeutics in hepatocellular carcinoma malignancies known to experience decreased CENP-E activity leading to aneuploidy. The predictions generated by computational models combined with experimental validations could further accelerate anticancer drug discovery. Citation Format: Pinyi Lu, Ryo Kamata, Michael R. Weil, Naomi Ohashi, Akihiro Ohashi, Eric A. Stahlberg. Computational modeling-based discovery of novel anticancer drug candidates targeting centromere-associated protein E [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4945.

AIoT-driven multi-source sensor emission monitoring and forecasting using multi-source sensor integration with reduced noise series decomposition

The integration of multi-source sensors based AIoT (Artificial Intelligence of Things) technologies into air quality measurement and forecasting is becoming increasingly critical in the fields of sustainable and smart environmental design, urban development, and pollution control. This study focuses on enhancing the prediction of emission, with a special emphasis on pollutants, utilizing advanced deep learning (DL) techniques. Recurrent neural networks (RNNs) and long short-term memory (LSTM) neural networks have shown promise in predicting air quality trends in time series data. However, challenges persist due to the unpredictability of air quality data and the scarcity of long-term historical data for training. To address these challenges, this study introduces the AIoT-enhanced EEMD-CEEMDAN-GCN model. This innovative approach involves decomposing the input signal using EEMD (Ensemble Empirical Mode Decomposition) and CEEMDAN (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) to extract intrinsic mode functions. These functions are then processed through a GCN (Graph Convolutional Network) model, enabling precise prediction of air quality trends. The model’s effectiveness is validated using air pollution datasets from four provinces in China, demonstrating its superiority over various deep learning models (GCN, EMD-GCN) and series decomposition models (EEMD-GCN, CEEMDAN-GCN). It achieves higher accuracy and better data fitting, outperforming other models in key metrics such as MAE (Mean Absolute Error), MSE (Mean Squared Error), MAPE (Mean Absolute Percentage Error), and R2 (Coefficient of Determination). The implementation of this AIoT-enhanced model in air pollution prediction allows decision-makers to more accurately anticipate changes in air quality, particularly concerning carbon emissions. This facilitates more effective planning of mitigation measures, improvement of public health, and optimization of resource allocation. Moreover, the model adeptly addresses the complexities of air quality data, contributing significantly to enhanced monitoring and management strategies in the context of sustainable urban development and environmental conservation.

Enhancing predictive models for illicit activities in the Bitcoin transaction network using advanced graph analytical techniques

The Elliptic dataset compiles a comprehensive history of Bitcoin transactions, integrating both anti-money laundering (AML) tags and distinct graph network features. Given the nature of the Bitcoin transaction networka complex, weakly interconnected structureleveraging graph analysis techniques for its study holds immense potential, especially in the realm of detecting illicit activities like hacking, drug trades, gambling, and more. A detailed examination of the Elliptic dataset, encompassing transaction amounts, frequencies, source and destination addresses, sheds light on the inherent structure and peculiarities of the Bitcoin transaction ecosystem. By conceptualizing this transactional landscape as a graph, a slew of analytical attributes emerge: node degree distribution, community architecture, centrality measures, and so forth. Such attributes pave the way for the creation of predictive models that can pinpoint and prognosticate potential unlawful trade actions. Several computational models have been employed on the Elliptic dataset, such as Logistic Regression (LR), Random Forest (RF), Multilayer Perceptrons (MLP), and Graph Convolutional Networks (GCN). The authors of this particular study delve into augmentations of the GCN model, juxtaposing the efficacy of the original GCN model against their enhanced algorithm within the context of the Elliptic dataset.

Drug Repositioning via Graph Neural Networks: Identifying Novel JAK2 Inhibitors from FDA-Approved Drugs through Molecular Docking and Biological Validation.

The increasing utilization of artificial intelligence algorithms in drug development has proven to be highly efficient and effective. One area where deep learning-based approaches have made significant contributions is in drug repositioning, enabling the identification of new therapeutic applications for existing drugs. In the present study, a trained deep-learning model was employed to screen a library of FDA-approved drugs to discover novel inhibitors targeting JAK2. To accomplish this, reference datasets containing active and decoy compounds specific to JAK2 were obtained from the DUD-E database. RDKit, a cheminformatic toolkit, was utilized to extract molecular features from the compounds. The DeepChem framework's GraphConvMol, based on graph convolutional network models, was applied to build a predictive model using the DUD-E datasets. Subsequently, the trained deep-learning model was used to predict the JAK2 inhibitory potential of FDA-approved drugs. Based on these predictions, ribociclib, topiroxostat, amodiaquine, and gefitinib were identified as potential JAK2 inhibitors. Notably, several known JAK2 inhibitors demonstrated high potential according to the prediction results, validating the reliability of our prediction model. To further validate these findings and confirm their JAK2 inhibitory activity, molecular docking experiments were conducted using tofacitinib-an FDA-approved drug for JAK2 inhibition. Experimental validation successfully confirmed our computational analysis results by demonstrating that these novel drugs exhibited comparable inhibitory activity against JAK2 compared to tofacitinib. In conclusion, our study highlights how deep learning models can significantly enhance virtual screening efforts in drug discovery by efficiently identifying potential candidates for specific targets such as JAK2. These newly discovered drugs hold promises as novel JAK2 inhibitors deserving further exploration and investigation.

Spatial-Pooling-Based Graph Attention U-Net for Hyperspectral Image Classification

In recent years, graph convolutional networks (GCNs) have attracted increasing attention in hyperspectral image (HSI) classification owing to their exceptional representation capabilities. However, the high computational requirements of GCNs have led most existing GCN-based HSI classification methods to utilize superpixels as graph nodes, thereby limiting the spatial topology scale and neglecting pixel-level spectral–spatial features. To address these limitations, we propose a novel HSI classification network based on graph convolution called the spatial-pooling-based graph attention U-net (SPGAU). Specifically, unlike existing GCN models that rely on fixed graphs, our model involves a spatial pooling method that emulates the region-growing process of superpixels and constructs multi-level graphs by progressively merging adjacent graph nodes. Inspired by the CNN classification framework U-net, SPGAU’s model has a U-shaped structure, realizing multi-scale feature extraction from coarse to fine and gradually fusing features from different graph levels. Additionally, the proposed graph attention convolution method adaptively aggregates adjacency information, thereby further enhancing feature extraction efficiency. Moreover, a 1D-CNN is established to extract pixel-level features, striking an optimal balance between enhancing the feature quality and reducing the computational burden. Experimental results on three representative benchmark datasets demonstrate that the proposed SPGAU outperforms other mainstream models both qualitatively and quantitatively.

Multiscale Feature Fusion and Graph Convolutional Network for Detecting Ethereum Phishing Scams

With the emergence of blockchain technology, the cryptocurrency market has experienced significant growth in recent years, simultaneously fostering environments conducive to cybercrimes such as phishing scams. Phishing scams on blockchain platforms like Ethereum have become a grave economic threat. Consequently, there is a pressing demand for effective detection mechanisms for these phishing activities to establish a secure financial transaction environment. However, existing methods typically utilize only the most recent transaction record when constructing features, resulting in the loss of vast amounts of transaction data and failing to adequately reflect the characteristics of nodes. Addressing this need, this study introduces a multiscale feature fusion approach integrated with a graph convolutional network model to detect phishing scams on Ethereum. A node basic feature set comprising 12 features is initially designed based on the Ethereum transaction dataset in the basic feature module. Subsequently, in the edge embedding representation module, all transaction times and amounts between two nodes are sorted, and a gate recurrent unit (GRU) neural network is employed to capture the temporal features within this transaction sequence, generating a fixed-length edge embedding representation from variable-length input. In the time trading feature module, attention weights are allocated to all embedding representations surrounding a node, aggregating the edge embedding representations and structural relationships into the node. Finally, combining basic and time trading features of the node, graph convolutional networks (GCNs), SAGEConv, and graph attention networks (GATs) are utilized to classify phishing nodes. The performance of these three graph convolution-based deep learning models is validated on a real Ethereum phishing scam dataset, demonstrating commendable efficiency. Among these, SAGEConv achieves an F1-score of 0.958, an AUC-ROC value of 0.956, and an AUC-PR value of 0.949, outperforming existing methods and baseline models.

GCN–Informer: A Novel Framework for Mid-Term Photovoltaic Power Forecasting

Predicting photovoltaic (PV) power generation is a crucial task in the field of clean energy. Achieving high-accuracy PV power prediction requires addressing two challenges in current deep learning methods: (1) In photovoltaic power generation prediction, traditional deep learning methods often generate predictions for long sequences one by one, significantly impacting the efficiency of model predictions. As the scale of photovoltaic power stations expands and the demand for predictions increases, this sequential prediction approach may lead to slow prediction speeds, making it difficult to meet real-time prediction requirements. (2) Feature extraction is a crucial step in photovoltaic power generation prediction. However, traditional feature extraction methods often focus solely on surface features, and fail to capture the inherent relationships between various influencing factors in photovoltaic power generation data, such as light intensity, temperature, and more. To overcome these limitations, this paper proposes a mid-term PV power prediction model that combines Graph Convolutional Network (GCN) and Informer models. This fusion model leverages the multi-output capability of the Informer model to ensure the timely generation of predictions for long sequences. Additionally, it harnesses the feature extraction ability of the GCN model from nodes, utilizing graph convolutional modules to extract feature information from the ‘query’ and ‘key’ components within the attention mechanism. This approach provides more reliable feature information for mid-term PV power prediction, thereby ensuring the accuracy of long sequence predictions. Results demonstrate that the GCN–Informer model significantly reduces prediction errors while improving the precision of power generation forecasting compared to the original Informer model. Overall, this research enhances the prediction accuracy of PV power generation and contributes to advancing the field of clean energy.

Identification of breast cancer subtypes based on graph convolutional network

Identification of molecular subtypes of malignant tumors plays a vital role in individualized diagnosis, personalized treatment, and prognosis prediction of cancer patients. The continuous improvement of comprehensive tumor genomics database and the ongoing breakthroughs in deep learning technology have driven further advancements in computer-aided tumor classification. Although the existing classification methods based on gene expression omnibus database take the complexity of cancer molecular classification into account, they ignore the internal correlation and synergism of genes. To solve this problem, we propose a multi-layer graph convolutional network model for breast cancer subtype classification combined with hierarchical attention network. This model constructs the graph embedding datasets of patients' genes, and develops a new end-to-end multi-classification model, which can effectively recognize molecular subtypes of breast cancer. A large number of test data prove the good performance of this new model in the classification of breast cancer subtypes. Compared to the original graph convolutional neural networks and two mainstream graph neural network classification algorithms, the new model has remarkable advantages. The accuracy, weight-F1-score, weight-recall, and weight-precision of our model in seven-category classification has reached 0.851 7, 0.823 5, 0.851 7 and 0.793 6 respectively. In the four-category classification, the results are 0.928 5, 0.894 9, 0.928 5 and 0.865 0 respectively. In addition, compared with the latest breast cancer subtype classification algorithms, the method proposed in this paper also achieved the highest classification accuracy. In summary, the model proposed in this paper may serve as an auxiliary diagnostic technology, providing a reliable option for precise classification of breast cancer subtypes in the future and laying the theoretical foundation for computer-aided tumor classification.

Spatial-temporal graph convolution network model with traffic fundamental diagram information informed for network traffic flow prediction

Accurate and fine-grained traffic state prediction has always been an important research field. For long-term traffic flow prediction, the high-dimensional and coupled traffic feature evolution patterns are deeply recessive, posing challenges in effectively characterizing and modeling them. This paper proposed a novel spatial–temporal graph convolution network model with traffic Fundamental Diagram (FD) information informed. The model decouples the high-dimensional spatiotemporal relationships in the transportation network and fully leverages the physical evolution laws of traffic states. First, the Graph Convolutional Network (GCN) with a spatial attention mechanism was proposed to capture spatial relations of the road network. The mechanism can better represent the spatial dynamics of the graph adjacency matrix in GCN. Second, this study injected prior physical knowledge into the graph adjacency matrix. This process was achieved by embedding characteristics of FDs from historical traffic data on the diagonal of the matrix, by which the propagation pattern of traffic states in road network could be considered. Third, to further catch the time dependence of the road network, the Gated Recurrent Unit (GRU) structure and the Transformer encoding structure were employed to locally and globally reform traffic state time sequences. Finally, experiments on a revised traffic dataset demonstrated that the proposed method consistently outperforms other baselines regarding Mean Absolute Error and Root Mean Square Error across all cases. Moreover, it achieved the optimal Mean Absolute Percentage Error in the 30- and 60-minute prediction tasks. This study shows a novel solution to inform traffic physical laws into data-driven state prediction models, and the reliability of the proposed method in long-term prediction offers valuable support for improving traffic management and alleviating traffic congestion.

An ultra-short-term wind power prediction method based on spatial-temporal attention graph convolutional model

Accurate and robust wind power forecasting (WPF) is great significance to ensure the safe and stable operation of the power system and promote the transformation of low-carbon energy. However, the high randomness and intermittency of wind power bring great challenges when designing reliable forecasting models. In this paper, a novel spatial-temporal attention graph convolutional network model is proposed. Firstly, the spatial attention mechanism is used to aggregate and extract the spatial correlations of the raw wind power data. Secondly, the temporal attention mechanism is applied to capture the temporal correlations. Then, the extracted spatial-temporal correlations were put into the temporal convolution network and the spatial convolution network to further obtain the temporal and spatial dependencies. Finally, the wind power forecasting results is output through the full connection layer. The proposed method is verified by using wind power data from real wind farm in China. The experimental results reveal that the proposed depth spatiotemporal prediction model has more significant advantages than other advanced models in terms of prediction accuracy and stability.

SEGCN: a subgraph encoding based graph convolutional network model for social bot detection

Message passing neural networks such as graph convolutional networks (GCN) can jointly consider various types of features for social bot detection. However, the expressive power of GCN is upper-bounded by the 1st-order Weisfeiler–Leman isomorphism test, which limits the detection performance for the social bots. In this paper, we propose a subgraph encoding based GCN model, SEGCN, with stronger expressive power for social bot detection. Each node representation of this model is computed as the encoding of a surrounding induced subgraph rather than encoding of immediate neighbors only. Extensive experimental results on two publicly available datasets, Twibot-20 and Twibot-22, showed that the proposed model improves the accuracy of the state-of-the-art social bot detection models by around 2.4%, 3.1%, respectively.

ScMGCN: A Multi-View Graph Convolutional Network for Cell Type Identification in scRNA-seq Data.

Single-cell RNA sequencing (scRNA-seq) data reveal the complexity and diversity of cellular ecosystems and molecular interactions in various biomedical research. Hence, identifying cell types from large-scale scRNA-seq data using existing annotations is challenging and requires stable and interpretable methods. However, the current cell type identification methods have limited performance, mainly due to the intrinsic heterogeneity among cell populations and extrinsic differences between datasets. Here, we present a robust graph artificial intelligence model, a multi-view graph convolutional network model (scMGCN) that integrates multiple graph structures from raw scRNA-seq data and applies graph convolutional networks with attention mechanisms to learn cell embeddings and predict cell labels. We evaluate our model on single-dataset, cross-species, and cross-platform experiments and compare it with other state-of-the-art methods. Our results show that scMGCN outperforms the other methods regarding stability, accuracy, and robustness to batch effects. Our main contributions are as follows: Firstly, we introduce multi-view learning and multiple graph construction methods to capture comprehensive cellular information from scRNA-seq data. Secondly, we construct a scMGCN that combines graph convolutional networks with attention mechanisms to extract shared, high-order information from cells. Finally, we demonstrate the effectiveness and superiority of the scMGCN on various datasets.

Multi-Channel Graph Convolutional Networks for Graphs with Inconsistent Structures and Features

Graph convolutional networks (GCNs) have attracted increasing attention in various fields due to their significant capacity to process graph-structured data. Typically, the GCN model and its variants heavily rely on the transmission of node features across the graph structure, which implicitly assumes that the graph structure and node features are consistent, i.e., they carry related information. However, in many real-world networks, node features may unexpectedly mismatch with the structural information. Existing GCNs fail to generalize to inconsistent scenarios and are even outperformed by models that ignore the graph structure or node features. To address this problem, we investigate how to extract representations from both the graph structure and node features. Consequently, we propose the multi-channel graph convolutional network (MCGCN) for graphs with inconsistent structures and features. Specifically, the MCGCN encodes the graph structure and node features using two specific convolution channels to extract two separate specific representations. Additionally, two joint convolution channels are constructed to extract the common information shared by the graph structure and node features. Finally, an attention mechanism is utilized to adaptively learn the importance weights of these channels under the guidance of the node classification task. In this way, our model can handle both consistent and inconsistent scenarios. Extensive experiments on both synthetic and real-world datasets for node classification and recommendation tasks show that our methods, MCGCN-A and MCGCN-I, achieve the best performance on seven out of eight datasets and the second-best performance on the remaining dataset. For simpler graph structures or tasks where the overhead of multiple convolution channels is not justified, traditional single-channel GCN models might be more efficient.

Early prediction of dementia using fMRI data with a graph convolutional network approach

Objective. Alzheimer’s disease is a progressive neurodegenerative dementia that poses a significant global health threat. It is imperative and essential to detect patients in the mild cognitive impairment (MCI) stage or even earlier, enabling effective interventions to prevent further deterioration of dementia. This study focuses on the early prediction of dementia utilizing Magnetic Resonance Imaging (MRI) data, using the proposed Graph Convolutional Networks (GCNs). Approach. Specifically, we developed a functional connectivity (FC) based GCN framework for binary classifications using resting-state fMRI data. We explored different types and processing methods of FC and evaluated the performance on the OASIS-3 dataset. We developed the GCN model for two different purposes: (1) MCI diagnosis: classifying MCI from normal controls (NCs); and (2) dementia risk prediction: classifying NCs from subjects who have the potential for developing MCI but have not been clinically diagnosed as MCI. Main results. The results of the experiments revealed several important findings: First, the proposed GCN outperformed both the baseline GCN and Support Vector Machine (SVM). It achieved the best average accuracy of 80.3% (11.7% higher than the baseline GCN and 23.5% higher than SVM) and the highest accuracy of 91.2%. Secondly, the GCN framework with (absolute) individual FC performed slightly better than that with global FC generally. However, GCN using global graphs with appropriate connectivity can achieve equivalent or superior performance to individual graphs in some cases, which highlights the significance of suitable connectivity for achieving performance. Additionally, the results indicate that the self-network connectivity of specific brain network regions (such as default mode network, visual network, ventral attention network and somatomotor network) may play a more significant role in GCN classification. Significance. Overall, this study offers valuable insights into the application of GCNs in brain analysis and early diagnosis of dementia. This contributes significantly to the understanding of MCI and has substantial potential for clinical applications in early diagnosis and intervention for dementia and other neurodegenerative diseases. Our code for GCN implementation is available at: https://github.com/Shuning-Han/FC-based-GCN.

Dynamic Graph Convolutional Network-Based Prediction of the Urban Grid-Level Taxi Demand–Supply Imbalance Using GPS Trajectories

The objective imbalance between the taxi supply and demand exists in various areas of the city. Accurately predicting this imbalance helps taxi companies with dispatching, thereby increasing their profits and meeting the travel needs of residents. The application of Graph Convolutional Networks (GCNs) in traffic forecasting has inspired the development of a spatial–temporal model for grid-level prediction of the taxi demand–supply imbalance. However, spatial–temporal GCN prediction models conventionally capture only static inter-grid correlation features. This research aims to address the dynamic influences caused by taxi mobility and the variations of other transportation modes on the demand–supply dynamics between grids. To achieve this, we employ taxi trajectory data and develop a model that incorporates dynamic GCN and Gated Recurrent Units (GRUs) to predict grid-level imbalances. This model captures the dynamic inter-grid influences between neighboring grids in the spatial dimension. It also identifies trends and periodic changes in the temporal dimension. The validation of this model, using taxi trajectory data from Shenzhen city, indicates superior performance compared to classical time-series models and spatial–temporal GCN models. An ablation study is conducted to analyze the impact of various factors on the predictive accuracy. This study demonstrates the precision and applicability of the proposed model.

TG468: a text graph convolutional network for predicting clinical response to immune checkpoint inhibitor therapy.

Enhancing cancer treatment efficacy remains a significant challenge in human health. Immunotherapy has witnessed considerable success in recent years as a treatment for tumors. However, due to the heterogeneity of diseases, only a fraction of patients exhibit a positive response to immune checkpoint inhibitor (ICI) therapy. Various single-gene-based biomarkers and tumor mutational burden (TMB) have been proposed for predicting clinical responses to ICI; however, their predictive ability is limited. We propose the utilization of the Text Graph Convolutional Network (GCN) method to comprehensively assess the impact of multiple genes, aiming to improve the predictive capability for ICI response. We developed TG468, a Text GCN model framing drug response prediction as a text classification task. By combining natural language processing (NLP) and graph neural network techniques, TG468 effectively handles sparse and high-dimensional exome sequencing data. As a result, TG468 can distinguish survival time for patients who received ICI therapy and outperforms single gene biomarkers, TMB and some classical machine learning models. Additionally, TG468's prediction results facilitate the identification of immune status differences among specific patient types in the Cancer Genome Atlas dataset, providing a rationale for the model's predictions. Our approach represents a pioneering use of a GCN model to analyze exome data in patients undergoing ICI therapy and offers inspiration for future research using NLP technology to analyze exome sequencing data.