Embedding-Based Entity Alignment of Cross-Lingual Temporal Knowledge Graphs

Entity alignment aims to construct a complete knowledge graph (KG) by matching the same entities in multi-source KGs. Existing researches on entity alignment mainly focuses on static multi-relational data in knowledge graphs. However, the relationships or attributes between entities often possess temporal characteristics as well. Neglecting these temporal characteristics can frequently lead to alignment errors. Compared to studying entity alignment in temporal knowledge graphs, there are relatively few efforts on entity alignment in cross-lingual temporal knowledge graphs. Therefore, in this paper, we put forward an entity alignment method for cross-lingual temporal knowledge graphs, namely CTEA. Based on GCN and TransE, CTEA combines entity embeddings, relation embeddings and attribute embeddings to design a joint embedding model, which is more conducive to generating transferable entity embedding. In the meantime, the distance calculation between elements and the similarity calculation of entity pairs are combined to enhance the reliability of cross-lingual entity alignment. Experiments shows that the proposed CTEA model improves Hits@m and MRR by about 0.8∼2.4 percentage points compared with the latest methods.

Explanatory subgraph attacks against Graph Neural Networks

Graph Neural Networks (GNNs) are often viewed as black boxes due to their lack of transparency, which hinders their application in critical fields. Many explanation methods have been proposed to address the interpretability issue of GNNs. These explanation methods reveal explanatory information about graphs from different perspectives. However, the explanatory information may also pose an attack risk to GNN models. In this work, we will explore this problem from the explanatory subgraph perspective. To this end, we utilize a powerful GNN explanation method called SubgraphX and deploy it locally to obtain explanatory subgraphs from given graphs. Then we propose methods for conducting evasion attacks and backdoor attacks based on the local explainer. In evasion attacks, the attacker gets explanatory subgraphs of test graphs from the local explainer and replace their explanatory subgraphs with an explanatory subgraph of other labels, making the target model misclassify test graphs as wrong labels. In backdoor attacks, the attacker employs the local explainer to select an explanatory trigger and locate suitable injection locations. We validate the effectiveness of our proposed attacks on state-of-art GNN models and different datasets. The results also demonstrate that our proposed backdoor attack is more efficient, adaptable, and concealed than previous backdoor attacks.

Brain Networks and Intelligence: A Graph Neural Network Based Approach to Resting State fMRI Data

Resting-state functional magnetic resonance imaging (rsfMRI) is a powerful tool for investigating the relationship between brain function and cognitive processes as it allows for the functional organization of the brain to be captured without relying on a specific task or stimuli. In this paper, we present a novel modeling architecture called BrainRGIN for predicting intelligence (fluid, crystallized and total intelligence) using graph neural networks on rsfMRI derived static functional network connectivity matrices. Extending from the existing graph convolution networks, our approach incorporates a clustering-based embedding and graph isomorphism network in the graph convolutional layer to reflect the nature of the brain sub-network organization and efficient network expression, in combination with TopK pooling and attention-based readout functions. We evaluated our proposed architecture on a large dataset, specifically the Adolescent Brain Cognitive Development Dataset, and demonstrated its effectiveness in predicting individual differences in intelligence. Our model achieved lower mean squared errors, and higher correlation scores than existing relevant graph architectures and other traditional machine learning models for all of the intelligence prediction tasks. The middle frontal gyrus exhibited a significant contribution to both fluid and crystallized intelligence, suggesting their pivotal role in these cognitive processes. Total composite scores identified a diverse set of brain regions to be relevant which underscores the complex nature of total intelligence.

Migrate demographic group for fair Graph Neural Networks

Graph Neural networks (GNNs) have been applied in many scenarios due to the superior performance of graph learning. However, fairness is always ignored when designing GNNs. As a consequence, biased information in training data can easily affect vanilla GNNs, causing biased results toward particular demographic groups (divided by sensitive attributes, such as race and age). There have been efforts to address the fairness issue. However, existing fair techniques generally divide the demographic groups by raw sensitive attributes and assume that are fixed. The biased information correlated with raw sensitive attributes will run through the training process regardless of the implemented fair techniques. It is urgent to resolve this problem for training fair GNNs. To tackle this problem, we propose a brand new framework, FairMigration, which is able to migrate the demographic groups dynamically, instead of keeping that fixed with raw sensitive attributes. FairMigration is composed of two training stages. In the first stage, the GNNs are initially optimized by personalized self-supervised learning, and the demographic groups are adjusted dynamically. In the second stage, the new demographic groups are frozen and supervised learning is carried out under the constraints of new demographic groups and adversarial training. Extensive experiments reveal that FairMigration achieves a high trade-off between model performance and fairness.

Artificial Intelligence in ADME Property Prediction

Absorption, distribution, metabolism, excretion (ADME) are key properties of a small molecule that govern pharmacokinetic profiles and impact its efficacy and safety. Computational methods such as machine learning and artificial intelligence have gained significant interest in both academic and industrial settings to predict pharmacokinetic properties of small molecules. These methods are applied in drug discovery to optimize chemical libraries, prioritize hits from biological screens, and optimize ADME properties of lead molecules. In the recent years, the drug discovery community witnessed the use of a range of neural network architectures such as deep neural networks, recurrent neural networks, graph neural networks, and transformer neural networks, which marked a paradigm shift in computer-aided drug design and development. This chapter discusses recent developments with an emphasis on their application to predict ADME properties.

Exploring the role of edge distribution in graph convolutional networks

Graph Convolutional Networks (GCNs) have shown remarkable performance in processing graph-structured data by leveraging neighborhood information for node representation learning. While most GCN models assume strong homophily within the networks they handle, some models can also handle heterophilous graphs. However, the selection of neighbors participating in the node representation learning process can significantly impact these models' performance. To address this, we investigate the influence of neighbor selection on GCN performance, focusing on the analysis of edge distribution through theoretical and empirical approaches. Based on our findings, we propose a novel GCN model called Graph Convolution Network with Improved Edge Distribution (GCN-IED). GCN-IED incorporates both direct edges, which rely on local neighborhood similarity, and hidden edges, obtained by aggregating information from multi-hop neighbors. We extensively evaluate GCN-IED on diverse graph benchmark datasets and observe its superior performance compared to other state-of-the-art GCN methods on heterophilous datasets. Our GCN-IED model, which considers the role of neighbors and optimizes edge distribution, provides valuable insights for enhancing graph representation learning and achieving superior performance on heterophilous graphs.

GNN-surv: Discrete-Time Survival Prediction Using Graph Neural Networks

Survival prediction models play a key role in patient prognosis and personalized treatment. However, their accuracy can be improved by incorporating patient similarity networks, which uncover complex data patterns. Our study uses Graph Neural Networks (GNNs) to enhance discrete-time survival predictions (GNN-surv) by leveraging relationships in these networks. We build these networks using cancer patients' genomic and clinical data and train various GNN models on them, integrating Logistic Hazard and PMF survival models. GNN-surv models exhibit superior performance in survival prediction across two urologic cancer datasets, outperforming traditional MLP models. They maintain robustness and effectiveness under varying graph construction hyperparameter μ values, with performance boosts of up to 14.6% and 7.9% in the time-dependent concordance index and reductions in the integrated brier score of 26.7% and 24.1% in the BLCA and KIRC datasets, respectively. Notably, these models also maintain their effectiveness across three different types of GNN models, suggesting potential adaptability to other cancer datasets. The superior performance of our GNN-surv models underscores their wide applicability in the fields of oncology and personalized medicine, providing clinicians with a more accurate tool for patient prognosis and personalized treatment planning. Future studies can further optimize these models by incorporating other survival models or additional data modalities.

Node injection for class-specific network poisoning

Graph Neural Networks (GNNs) are powerful in learning rich network representations that aid the performance of downstream tasks. However, recent studies showed that GNNs are vulnerable to adversarial attacks involving node injection and network perturbation. Among these, node injection attacks are more practical as they do not require manipulation in the existing network and can be performed more realistically. In this paper, we propose a novel problem statement - a class-specific poison attack on graphs in which the attacker aims to misclassify specific nodes in the target class into a different class using node injection. Additionally, nodes are injected in such a way that they camouflage as benign nodes. We propose NICKI, a novel attacking strategy that utilizes an optimization-based approach to sabotage the performance of GNN-based node classifiers. NICKI works in two phases - it first learns the node representation and then generates the features and edges of the injected nodes. Extensive experiments and ablation studies on four benchmark networks show that NICKI is consistently better than four baseline attacking strategies for misclassifying nodes in the target class. We also show that the injected nodes are properly camouflaged as benign, thus making the poisoned graph indistinguishable from its clean version w.r.t various topological properties.

LoyalDE: Improving the performance of Graph Neural Networks with loyal node discovery and emphasis

Recent years have witnessed an increasing focus on graph-based semi-supervised learning with Graph Neural Networks (GNNs). Despite existing GNNs having achieved remarkable accuracy, research on the quality of graph supervision information has inadvertently been ignored. In fact, there are significant differences in the quality of supervision information provided by different labeled nodes, and treating supervision information with different qualities equally may lead to sub-optimal performance of GNNs. We refer to this as the graph supervision loyalty problem, which is a new perspective for improving the performance of GNNs. In this paper, we devise FT-Score to quantify node loyalty by considering both the local feature similarity and the local topology similarity, and nodes with higher loyalty are more likely to provide higher-quality supervision. Based on this, we propose LoyalDE (Loyal Node Discovery and Emphasis), a model-agnostic hot-plugging training strategy, which can discover potential nodes with high loyalty to expand the training set, and then emphasize nodes with high loyalty during model training to improve performance. Experiments demonstrate that the graph supervision loyalty problem will fail most existing GNNs. In contrast, LoyalDE brings about at most 9.1% performance improvement to vanilla GNNs and consistently outperforms several state-of-the-art training strategies for semi-supervised node classification.

Explainable Graph Neural Networks for Atherosclerotic Cardiovascular Disease

Understanding the aspects of progression for atherosclerotic cardiovascular disease and treatment is key to building reliable clinical decision-support systems. To promote system trust, one step is to make the machine learning models (used by the decision support systems) explainable for clinicians, developers, and researchers. Recently, working with longitudinal clinical trajectories using Graph Neural Networks (GNNs) has attracted attention among machine learning researchers. Although GNNs are seen as black-box methods, promising explainable AI (XAI) methods for GNNs have lately been proposed. In this paper, which describes initial project stages, we aim at utilizing GNNs for modeling, predicting, and exploring the model explainability of the low-density lipoprotein cholesterol level in long-term atherosclerotic cardiovascular disease progression and treatment.

A systematic review of biologically-informed deep learning models for cancer: fundamental trends for encoding and interpreting oncology data

BACKGROUND

There is an increasing interest in the use of Deep Learning (DL) based methods as a supporting analytical framework in oncology. However, most direct applications of DL will deliver models with limited transparency and explainability, which constrain their deployment in biomedical settings.

METHODS

This systematic review discusses DL models used to support inference in cancer biology with a particular emphasis on multi-omics analysis. It focuses on how existing models address the need for better dialogue with prior knowledge, biological plausibility and interpretability, fundamental properties in the biomedical domain. For this, we retrieved and analyzed 42 studies focusing on emerging architectural and methodological advances, the encoding of biological domain knowledge and the integration of explainability methods.

RESULTS

We discuss the recent evolutionary arch of DL models in the direction of integrating prior biological relational and network knowledge to support better generalisation (e.g. pathways or Protein-Protein-Interaction networks) and interpretability. This represents a fundamental functional shift towards models which can integrate mechanistic and statistical inference aspects. We introduce a concept of bio-centric interpretability and according to its taxonomy, we discuss representational methodologies for the integration of domain prior knowledge in such models.

CONCLUSIONS

The paper provides a critical outlook into contemporary methods for explainability and interpretability used in DL for cancer. The analysis points in the direction of a convergence between encoding prior knowledge and improved interpretability. We introduce bio-centric interpretability which is an important step towards formalisation of biological interpretability of DL models and developing methods that are less problem- or application-specific.

SymptomGraph: Identifying Symptom Clusters from Narrative Clinical Notes using Graph Clustering

Patients with cancer or other chronic diseases often experience different symptoms before or after treatments. The symptoms could be physical, gastrointestinal, psychological, or cognitive (memory loss), or other types. Previous research focuses on understanding the individual symptoms or symptom correlations by collecting data through symptom surveys and using traditional statistical methods to analyze the symptoms, such as principal component analysis or factor analysis. This research proposes a computational system, SymptomGraph, to identify the symptom clusters in the narrative text of written clinical notes in electronic health records (EHR). SymptomGraph is developed to use a set of natural language processing (NLP) and artificial intelligence (AI) methods to first extract the clinician-documented symptoms from clinical notes. Then, a semantic symptom expression clustering method is used to discover a set of typical symptoms. A symptom graph is built based on the co-occurrences of the symptoms. Finally, a graph clustering algorithm is developed to discover the symptom clusters. Although SymptomGraph is applied to the narrative clinical notes, it can be adapted to analyze symptom survey data. We applied Symptom-Graph on a colorectal cancer patient with and without diabetes (Type 2) data set to detect the patient symptom clusters one year after the chemotherapy. Our results show that SymptomGraph can identify the typical symptom clusters of colorectal cancer patients' post-chemotherapy. The results also show that colorectal cancer patients with diabetes often show more symptoms of peripheral neuropathy, younger patients have mental dysfunctions of alcohol or tobacco abuse, and patients at later cancer stages show more memory loss symptoms. Our system can be generalized to extract and analyze symptom clusters of other chronic diseases or acute diseases like COVID-19.

Patient-GAT: Sarcopenia Prediction using Multi-modal Data Fusion and Weighted Graph Attention Networks

Graph Attention Networks (GAT) have been extensively used to perform node-level classification on data that can be represented as a graph. However, few papers have investigated the effectiveness of using GAT on graph representations of patient similarity networks. This paper proposes Patient-GAT, a novel method to predict chronic health conditions by first integrating multi-modal data fusion to generate patient vector representations using imputed lab variables with other structured data. This data representation is then used to construct a patient network by measuring patient similarity, finally applying GAT to the patient network for disease prediction. We demonstrated our framework by predicting sarcopenia using real-world EHRs obtained from the Indiana Network for Patient Care. We evaluated the performance of our system by comparing it to other baseline models, showing that our model outperforms other methods. In addition, we studied the contribution of the temporal representation of the lab data and discussed the interpretability of this model by analyzing the attention coefficients of the trained Patient-GAT model. Our code can be found on Github.

Path reliability-based graph attention networks

Self-attention mechanism has been successfully introduced in Graph Neural Networks (GNNs) for graph representation learning and achieved state-of-the-art performances in tasks such as node classification and node attacks. In most existing attention-based GNNs, attention score is only computed between two directly connected nodes with their representation at a single layer. However, this attention score computation method cannot account for its multi-hop neighbors, which supply graph structure information and have influence on many tasks such as link prediction, knowledge graph completion, and adversarial attack as well. In order to address this problem, in this paper, we propose Path Reliability-based Graph Attention Networks (PRGATs), a novel method to incorporate multi-hop neighboring context into attention score computation, enabling to capture longer-range dependencies and large-scale structural information within a single layer. Moreover, path reliability-based attention layer, a core layer of PRGATs, uses a resource-constrain allocation algorithm to compute the reliable path and its attention scores from neighboring nodes to non-neighboring nodes, increasing the receptive field for every message-passing layer. Experimental results on real-world datasets show that, as compared with baselines, our model outperforms existing methods up to 3% on standard node classification and 12% on graph universal adversarial attack.

Multi-granularity heterogeneous graph attention networks for extractive document summarization

Extractive document summarization is a fundamental task in natural language processing (NLP). Recently, several Graph Neural Networks (GNNs) are proposed for this task. However, most existing GNN-based models can neither effectively encode semantic nodes of multiple granularity level apart from sentences nor substantially capture different cross-sentence meta-paths. To address these issues, we propose MHgatSum, a novel Multi-granularity Heterogeneous Graph ATtention networks for extractive document SUMmarization. Specifically, we first build a multi-granularity heterogeneous graph (HetG) for each document, which is better to represent the semantic meaning of the document. The HetG contains not only sentence nodes but also multiple other granularity effective semantic units with different semantic levels, including keyphrases and topics. These additional nodes act as the intermediary between sentences to build the meta-paths involved in sentence node (i.e., Sentence-Keyphrase-Sentence and Sentence-Topic-Sentence). Then, we propose a heterogeneous graph attention networks to embed the constructed HetG for extractive summarization, which enjoys multi-granularity semantic representations. The model is based on a hierarchical attention mechanism, including node-level and semantic-level attentions. The node-level attention can learn the importance between a node and its meta-path based neighbors, while the semantic-level attention is able to learn the importance of different meta-paths. Moreover, to better integrate sentence global knowledge, we further incorporate sentence node global importance in local node-level attention. We conduct empirical experiments on two benchmark datasets, which demonstrates the superiority of MHgatSum over previous SOTA models on the task of extractive summarization.

Structure-Aware Siamese Graph Neural Networks for Encounter-Level Patient Similarity Learning

Patient similarity learning has attracted great research interest in biomedical informatics. Correctly identifying the similarity between a given patient and patient records in the database could contribute to clinical references for diagnosis and medication. The sparsity of underlying relationships between patients poses difficulties for similarity learning, which becomes more challenging when considering real-world Electronic Health Records (EHRs) with a large number of missing values. In the paper, we organize EHRs as a graph and propose a novel deep learning framework, Structure-aware Siamese Graph neural Networks (SSGNet), to perform robust encounter-level patient similarity learning while capturing the intrinsic graph structure and mitigating the influence from missing values. The proposed SSGNet regards each patient encounter as a node, and learns the node embeddings and the similarity between nodes simultaneously via Graph Neural Networks (GNNs) with siamese architecture. Further, SSGNet employs a low-rank and contrastive objective to optimize the structure of the patient graph and enhance model capacity. The extensive experiments were conducted on two publicly available datasets and a real-world dataset regarding IgA nephropathy from Peking University First Hospital, in comparison with multiple baseline and state-of-the-art methods. The significant improvement in Accuracy, Precision, Recall and F1 score on the patient encounter pairwise similarity classification task demonstrates the superiority of SSGNet. The mean average precision (mAP) of SSGNet on the similar encounter retrieval task is also better than other competitors. Furthermore, SSGNet's stable similarity classification accuracies at different missing rates of data validate the effectiveness and robustness of our proposal.

Geometric graphs from data to aid classification tasks with Graph Convolutional Networks

Traditional classification tasks learn to assign samples to given classes based solely on sample features. This paradigm is evolving to include other sources of information, such as known relations between samples. Here, we show that, even if additional relational information is not available in the dataset, one can improve classification by constructing geometric graphs from the features themselves, and using them within a Graph Convolutional Network. The improvement in classification accuracy is maximized by graphs that capture sample similarity with relatively low edge density. We show that such feature-derived graphs increase the alignment of the data to the ground truth while improving class separation. We also demonstrate that the graphs can be made more efficient using spectral sparsification, which reduces the number of edges while still improving classification performance. We illustrate our findings using synthetic and real-world datasets from various scientific domains.

•

Geometric graphs from data can be used in deep learning to improve classification

•

Optimized graphs align the data to the class labels and enhance class separability

•

Sparsifying the optimized graph can potentially improve classification performance

•

Extensive experiments are performed on datasets from various scientific domains

Supervised classification assigns unseen samples to classes based on their features by learning from examples with known class labels. We show that classification can be improved by using the sample features not only as the basis for classification, but also as a means to construct geometric graphs that encapsulate the closeness between samples. Such feature-derived graphs can be used within graph-based deep-learning models to improve classification. To understand the benefits of these graphs, we show that they align the data to the class labels and enhance class separability. We also demonstrate how to make the graphs sparser, and hence more efficient, while still potentially improving their performance. Our findings are timely given the increasing interest in combining graphs with classification and learning tasks.