Causal Network Research Articles

Tutorial on Multiple Mediation Analysis Using Causal Networks: Application to Diagnosing COVID-19 From Its Early and Late Symptoms.

There are two methods of studying multiple mediation: network-based and analysis of coefficients in regression equations. 822 individuals who had completed a COVID-19 test were recruited through listservs and via employees and patients of Virginia Commonwealth University Health Center. Participants reported their symptoms and which symptom(s) occurred first. A Causal Network model was established through a repeated chain of regressions in four steps: First, we identified the order of occurrence of symptoms. Second, COVID-19 test results were LASSO regressed on symptoms and demographic variables, establishing direct effects. Third, the direct effects were LASSO regressed on prior symptoms and demographic variables, establishing indirect effects. Fourth, the joint distribution of the variables in the network was simulated by evaluating regression equations at factorial combinations of their direct effects. Fifth, the mediated effect was calculated through twin modeling, where the model derived from the real data was compared to the counterfactual model that represented 'what if' there was no mediation. The 10-fold cross-validated area under the receiver curve for the network model was 0.82, which is a moderate to high level of accuracy. The network model identified later symptoms (e.g., chills) mediated the effect of earlier symptoms (e.g. fever). A network-based multiple mediation analysis led to new insights by integrating findings of 19 separate regressions into a single network model. The procedure showed how artificial intelligence can help in triage of COVID-19 patients from their symptoms, before any home or laboratory tests.

Updating patient perceptions with intensive longitudinal data for enhanced case conceptualizations: An approach with Bayesian informative priors.

Addressing the persistent heterogeneity in psychopathology, treatment outcomes, and the science-practice gap requires a systematic approach to personalizing psychotherapy. Case conceptualization seeks to understand a patient's unique psychopathology by generating and continuously updating hypotheses about predisposing, precipitating, and maintaining factors. This study introduces a new data-driven method to formalize this process with personalized network estimation, combining prior elicitation and Bayesian inference. It is the first to test its clinical usefulness with 12 patients, primarily treated for depression, and their therapists (preregistered and can be found as the additional online materials: https://osf.io/38qdx). Patients employed the Perceived Causal Networks method to create personalized "prior networks," mapping how they perceived their symptoms to interact. Bayesian inference was used to update these prior networks using longitudinal data collected subsequently 6 times daily over 15 days (N = 935), resulting in personalized "posterior networks." Both Perceived Causal Networks and longitudinal assessments were evaluated as feasible and acceptable. Face validity was scored highest for the posterior networks. Patients emphasized the personal relevance of these networks, while therapists noted their value in guiding the therapeutic process. However, prior, posterior, and data networks showed significant dissimilarities. These differences may stem from patients' limited insight into symptom interactions, insufficient power in the longitudinal data, or variations in self-perception. Despite some inconsistencies, the study shows potential for combining two methods to create personalized models of psychopathology, highlighting the need for future research to refine this formalization process into a more rigorous theoretical-empirical cycle to test these models. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Exploring the Essence of Relationships for Scene Graph Generation via Causal Features Enhancement Network.

Scene graph generation (SGG) establishes a structured representation between multiple objects by exploring their relationship for visual perception and reasoning tasks. Existing SGG methods often fit the relationships' distribution by introducing language prior or statistical knowledge. However, the relationships should be the semantic reflection of the interaction between objects, rather than the statistical dependency between their categories. To solve this problem, we propose a novel Causal Features Enhancement Network (CFEN) to mine the essential semantic features between objects and relationships. Specifically, by decomposing the object features into class-generic and object-specific components, the causal graph framework is designed to analyze these existing SGG methods. To measure the influence of object-specific features for relationship recognition, we construct the counterfactual training framework for computing the difference between fact and counterfactual logits. Besides, to strengthen the role of object-specific features and learn the interaction between objects, a distribution matching loss is proposed to compute the KL divergence between counterfactual outputs and standard difference distributions and modulate the relations predictions. Finally, compared with the current state-of-the-art methods, the extensive experimental results on VG150 and VrR-VG datasets demonstrate the effectiveness and superiority of our proposed CFEN.

Effective connectivity analysis of response inhibition functional network

IntroductionInhibition mechanisms are essential in daily life, helping individuals adapt to environmental demands. However, the causal interactions between large-scale functional networks involved in response inhibition remain poorly understood.MethodsIn this study, we examined the effective connectivity (EC) underlying inhibitory processes in the brain using dynamic causal modeling (DCM) and independent component analysis (ICA). We conducted a Go-NoGo fMRI task with 19 healthy participants to investigate these networks.ResultsOur results identified four functional networks activated during correct motor response inhibition: the salience network (SN), the right and left executive control networks (ECNs), and the ventral default mode network (vDMN). We observed a significant causal inhibitory influence from the vDMN to the left ECN (lECN). Under conditions of unsuccessful response inhibition, the SN, bilateral ECNs, and somatomotor network (SMN) were found to be prominently activated. Furthermore, we identified a significant correlation between the inhibitory influence from the SMN to the SN and the commission error rate. Finally, correlation analyses between self-reported impulsivity levels and causal network interactions revealed that highly impulsive individuals require greater interhemispheric integration between the right and left ECNs for effective inhibition, as well as a causal excitatory modulation from the right executive control network (rECN) to the vDMN.DiscussionIn summary, our study reveals complex hierarchical dynamics among functional networks during response inhibition. These findings offer valuable insight into the neural mechanisms supporting inhibition and provide avenues for future research on the neural underpinnings of this critical cognitive function across the lifespan.

Causal network inference of cis- and trans-gene regulation of expression quantitative trait loci across human tissues

Abstract Expression quantitative trait loci (eQTLs) have been identified for most genes in the human genome across tissues and cell types. While most of the eQTLs are near (i.e., cis) the associated genes on the linear genome, some can be far away or on different chromosomes (i.e., trans), with the regulatory mechanisms largely unknown. Here, we study regulation by eQTLs of their cis- and trans-genes across nearly 50 tissues and cell types, taking a causal network inference approach and leveraging the principle of Mendelian randomization. Specifically, we constructed trios consisting of an eQTL, its cis-gene and trans-gene. We then inferred the regulatory relationships, using the eQTL as an instrumental variable and accounting for confounding variables. We identify multiple types of regulatory networks for trios: across all the tissues, more than half of the trios are inferred to be conditionally independent, where the two genes are conditionally independent given the genotype of the eQTL (gene 1 ← eQTL → gene 2). Around 1.5% of the trios are inferred to be mediation (eQTL → mediator → target), around 1.3% fully connected among the three nodes, and just a handful v-structures (eQTL → gene 1 ← gene 2). The identifications are generally consistent with the statistical dependence patterns of each trio. Genes in trios of different regulatory types exhibit distinct functional enrichments. Interestingly, many mediation trios have the trans-gene as the mediator. Existing transcription factor databases and HiC data for genome spatial structure provide additional support for long-range cis-acting and trans-acting in some of the inferred trans-gene mediation trios.

Progressive gray matter alterations in the Meige's syndrome and across sub-types.

Progressive gray matter alterations in the Meige's syndrome and across sub-types.

HCN-RLR-CAN: A novel human-computer negotiation model based on round-level recurrence and causal attention networks

HCN-RLR-CAN: A novel human-computer negotiation model based on round-level recurrence and causal attention networks

DSAM: A deep learning framework for analyzing temporal and spatial dynamics in brain networks.

DSAM: A deep learning framework for analyzing temporal and spatial dynamics in brain networks.

Root Cause Diagnosis for Time Series Data Faults Based on Fault Isolation and Channel-Attentive Sparse Causal Networks

Root Cause Diagnosis for Time Series Data Faults Based on Fault Isolation and Channel-Attentive Sparse Causal Networks

MR.RGM: an R package for fitting Bayesian multivariate bidirectional Mendelian randomization networks.

Mendelian randomization (MR) infers causal relationships between exposures and outcomes using genetic variants as instrumental variables. Typically, MR considers only a pair of exposure and outcome at a time, limiting its capability of capturing the entire causal network. We overcome this limitation by developing MR.RGM (Mendelian randomization via reciprocal graphical model), a fast R-package that implements the Bayesian reciprocal graphical model and enables practitioners to construct holistic causal networks with possibly cyclic/reciprocal causation and proper uncertainty quantifications, offering a comprehensive understanding of complex biological systems and their interconnections. We developed MR.RGM, an open-source R package that applies bidirectional MR using a network-based strategy, enabling the exploration of causal relationships among multiple variables in complex biological systems. MR.RGM holds the promise of unveiling intricate interactions and advancing our understanding of genetic networks, disease risks, and phenotypic complexities. MR.RGM is available at CRAN (https://CRAN.R-project.org/package=MR.RGM, DOI: 10.32614/CRAN.package.MR.RGM) and https://github.com/bitansa/MR.RGM.

Construction of Knowledge Graph for Marine Diesel Engine Faults Based on Deep Learning Methods

As the core equipment in ship power systems, the accurate and real-time diagnosis of ship diesel engine faults directly affects navigation safety and operation efficiency. Existing methods (e.g., expert systems, traditional machine learning) can hardly cope with the complex failure modes and dynamic operation environment due to the problems of relying on artificial features and insufficient generalization ability. In this paper, we propose a BiLSTM-CRF-based knowledge graph construction method for ship diesel engine faults, aiming at integrating multi-source heterogeneous data through deep learning and knowledge graph technology, and mining the deep semantic associations among fault phenomena, causes, and solutions. The research framework covers data acquisition, ontology modeling, and knowledge extraction and storage, and the BiLSTM-CRF model is used to fuse bi-directional contextual features with label transfer probability to achieve high-precision entity recognition and relationship extraction. Finally, a scalable knowledge graph is constructed by Neo4j. Experiments show that the model significantly outperforms baseline methods such as HMM, CRF, and BiLSTM, and the graph visualization clearly presents the fault causality network, which supports knowledge reasoning and decision optimization. For example, “high exhaust temperature” can be related to potential causes such as “turbine failure” and “poor cooling”, and recommended measures can be taken. This method not only improves fault diagnosis accuracy and efficiency but also provides a novel method for intelligent ship health management.

36 Kidney MiRNA expression in BTBR ob/ob mice at a critical time point in disease development and progression

Objectives/Goals: Diabetic kidney disease (DKD) affects 40% of diabetic patients, leading to renal failure, yet the molecular drivers remain elusive. MicroRNAs, noncoding regulators of gene expression, may hold the key. This study aims to identify key miRNAs in DKD, providing crucial insights for early intervention. Methods/Study Population: miRNA sequencing was conducted on kidneys from 8-week old male BTBR wild type and BTBR ob/ob mice. BTBR ob/ob mice lack the hormone leptin and spontaneously develop type 2 diabetes, with morphological renal lesions characteristic of human DKD. Total RNA was extracted from whole kidney sections and processed using the QIAseq miRNA library kit. Sequencing was performed on an Illumina NextSeq 550 platform. GeneGlobe analysis was used to identify differentially expressed miRNA functional pathways, while ingenuity pathway analysis (IPA) was employed to predict master regulators and causal networks involved in DKD. Results/Anticipated Results: miRNA sequencing identified significantly differentially expressed miRNAs (p < 0.05) between 8-week-old BTBR WT and BTBR ob/ob male mice, including miR-34a (-6.86 fold), miR-122 (-5.01 fold), miR-129 (-2.23 fold), miR-142a (+2.78 fold), miR-346 (+4.66 fold), miR-547 (-2.49 fold), miR-592 (+11.81 fold), miR-802 (-6.95 fold), and miR-6539 (-7.93 fold). Qiagen GeneGlobe analysis revealed biological processes potentially targeted by these miRNAs, including endocytosis, phagocytosis, hyperglycemia (p = 7.59e-3), and insulin-dependent diabetes (p = 4.32e-4). IPA predicted activation of RRAS, a small GTPase regulating cell growth and signaling (Z-score +2), with miR-34a and miR-122 targeting MYC, PI3K, and TGF-β in DKD progression in BTBR ob/ob mice. Discussion/Significance of Impact: We identified kidney miRNA expression in BTBR ob/ob mice at a pivotal disease stage. miR-34a, miR-122, and RRAS emerged as key drivers in DKD progression, showing remarkable early biomarker potential. These findings lay the groundwork for early detection and innovative therapies to halt DKD and improve patient outcomes.

Multi-trait phenotypic modeling through factor analysis and bayesian network learning to develop latent reproductive, body conformational, and carcass-associated traits in admixed beef heifers.

Despite high-throughput and large-scale phenotyping becoming easier, interpretation of such data in cattle production remains challenging due to the complex and highly correlated nature of many traits. Underlying biological traits (UBT) of economic importance are defined by a subset of easy-to-measure traits, leading to challenges in making appropriate selection decisions on them. Research on UBT in beef cattle is limited. In this study, the phenotypic data of admixed beef heifers (n = 336) for reproductive, body conformation, and carcass-related traits (traits, t = 35) were used to identify latent variables from factor analysis (FA) that can be characterized as UBT. Given sample size constraints for carcass (n = 161) and other body size-related traits (n = 336), two models were explored. In Model 1, all individual traits were considered (n = 161), while in Model 2, the dataset was split into body size (n = 336) and carcass (n = 161) traits to maximize available heifers per dataset. A combination of FA and Bayesian network (BN) learning was adopted to develop UBT and infer BN structure for subsequent analyses. All heifers (n = 336) were genotyped using GeneSeek Genomic Profiler 150K for Beef Cattle. Following quality checks, 117,373 autosomal SNP markers were retained and used for genomic estimated breeding values (gEBV) in BN learning steps. Using exploratory and confirmatory FA, Body Size (BS) and Body Composition (BC) were identified as UBT for Model 1, explaining 14 phenotypic traits (t = 14). For Model 2, BS, Ovary Size, and Yield Grade (YG) were identified as UBT, explaining 12 phenotypic traits (t = 12). When using gEBV, the causal network structure inferred showed BS contributed to BC in Model 1 and to Ovary Size in Model 2. Therefore, a structure equation-based approach should be used in subsequent modeling for these traits. From Model 2, YG should be modeled univariately. This study is the first to identify UBT in growing admixed heifers using body size, conformation, and carcass traits. We also identified that BC and YG did not explain intra-muscular fat and body density, indicating these two traits should also be modeled univariately.

Temporal changes in the protein cargo of extracellular vesicles and resultant immune reprogramming after severe burn injury in humans and mice.

Severe injury, including burn trauma, leads to profound immune dysfunction, yet the mechanisms driving these changes remain incompletely defined. This lack of understanding has hindered efforts to modulate the immune response effectively. Additionally, a clear biomarker profile to guide clinicians in identifying burn patients at high risk for poor clinical outcomes is lacking. Extracellular vesicles (EVs) have emerged as novel mediators of immune dysfunction in various pathologies. Prior studies in mouse models have demonstrated that plasma EVs increase following burn injury and contribute to immune dysfunction. Furthermore, EVs have potential as biomarkers for predicting extended hospital stays in burn patients. This study hypothesizes that human EVs, purified early and late after burn injury, will exhibit immune reprogramming effects similar to those observed in mice and that specific EV protein cargo may serve as biomarkers of immune and physiological responses to burn injury. EVs were isolated from the plasma of burn-injury patients at early (<72h) and late (≥14 days) time points post-injury. Using unbiased immune transcriptome and bioinformatic causal network analyses, the immunomodulatory effects of these EVs were assessed in human THP-1 macrophages. Mass spectrometry-based quantitative proteomics and pathway analyses were conducted to characterize the protein cargo of EVs from both human and mouse models at different post-burn phases. Early post-burn human EVs induced significant immune reprogramming in macrophages, increasing pro-inflammatory signaling while suppressing anti-inflammatory pathways. In contrast, late post-burn EVs exhibited an immunosuppressive profile, with downregulation of pro-inflammatory pathways and upregulation of anti-inflammatory signaling. Proteomic analyses revealed that human and mouse EVs contained unique and overlapping protein cargo across different time points. At day 7 post-burn, mouse EVs were enriched in circulation/complement and neuronal proteins, whereas by day 14, reductions in membrane and metabolism-associated proteins were observed. Similarly, in human EVs at 14 days post-burn, increased levels of circulation/complement, immune, and transport proteins were detected. EVs from burn-injury patients at distinct time points differentially modulate immune responses in macrophages, mirroring the temporal immune phenotypes observed in clinical settings. These findings suggest that EV-macrophage interactions play a crucial role in burn-induced immune dysfunction and highlight the potential of EV protein cargo as biomarkers for immune status and patient outcomes following burn injury. Human extracellular vesicles released into the plasma after severe burn injury can reprogram the immune system with corresponding immunomodulatory protein cargo.

Hierarchical Regulatory Networks Reveal Conserved Drivers of Plant Drought Response at the Cell-Type Level.

Drought is a critical environmental challenge affecting plant growth and productivity. Understanding the regulatory networks governing drought response at the cellular level remains an open question. Here, a comprehensive multi-omics integration framework that combines transcriptomic, proteomic, epigenetic, and network-based analyses to delineate cell-type-specific regulatory networks involved in plant drought response is presented. By analyzing nearly 30000 multi-omics data samples across species, unique insights are revealed into conserved drought responses and cell-type-specific regulatory dynamics, leveraging novel integrative analytical workflows. Notably, CIPK23 emerges as a conserved protein kinase mediating drought tolerance through interactions with CBL4, as validated by yeast two-hybrid and BiFC assays. Experimental validation in Arabidopsis thaliana and Vitis vinifera confirms the functional conservation of CIPK23, which enhances drought resistance in overexpression lines. In addition, the authors' causal network analysis pinpoints critical regulatory drivers such as NLP7 and CIPK23, providing insights into the molecular mechanisms of drought adaptation. These findings advance understanding of plant drought tolerance and offer potential targets for improving crop resilience across diverse species.

A large-scale benchmark for network inference from single-cell perturbation data

Mapping biological mechanisms in cellular systems is a fundamental step in early-stage drug discovery that serves to generate hypotheses on what disease-relevant molecular targets may effectively be modulated by pharmacological interventions. With the advent of high-throughput methods for measuring single-cell gene expression under genetic perturbations, we now have effective means for generating evidence for causal gene-gene interactions at scale. However, evaluating the performance of network inference methods in real-world environments is challenging due to the lack of ground-truth knowledge. Moreover, traditional evaluations conducted on synthetic datasets do not reflect the performance in real-world systems. We thus introduce CausalBench, a benchmark suite revolutionizing network inference evaluation with real-world, large-scale single-cell perturbation data. CausalBench, distinct from existing benchmarks, offers biologically-motivated metrics and distribution-based interventional measures, providing a more realistic evaluation of network inference methods. An initial systematic evaluation of state-of-the-art causal inference methods using our CausalBench suite highlights how poor scalability of existing methods limits performance. Moreover, methods that use interventional information do not outperform those that only use observational data, contrary to what is observed on synthetic benchmarks. CausalBench subsequently enables the development of numerous promising methods through a community challenge, thus demonstrating its potential as a transformative tool in the field of computational biology, bridging the gap between theoretical innovation and practical application in drug discovery and disease understanding. Thus, CausalBench opens new avenues for method developers in causal network inference research, and provides to practitioners a principled and reliable way to track progress in network methods for real-world interventional data.

Causal interaction of metabolic oscillations in monolayers of Hela cervical cancer cells: emergence of complex networks

A novel global cooperative phenomenon was observed in monolayers of HeLa cervical cancer cells that exhibited glycolytic oscillations but did not exhibit synchronisation or partial synchronisation. The analysis of causality of the oscillations between cell pairs in the cell-monolayer sheet revealed a hidden causal interaction network. Furthermore, the network exhibits characteristics of a broad-scale network. This suggests that key cells perform a hub-like function in the network and that the population of HeLa cells forms metabolically connected functional network rather than randomly connected one. Unlike previous work that focused on the synchronisation of glycolytic oscillations in the HeLa cells, the present study analysed the causality between the oscillating cells by Convergent Cross Mapping (CCM), which is based on the phase-space reconstruction of time-series data and is used to find causality in weakly coupled components of nonlinear dynamical systems. We believe that the framework proposed in this study is useful for investigating the hidden state of a group of cells and can accelerate studies on cellular metabolic phenomena including metabolic oscillations in cells within islets of Langerhans. It would also be applicable to systems of weakly coupled oscillators that may include hidden cooperative phenomena.

Obesity is associated with progressive brain structural changes.

The aim of this study was to investigate the relationship between obesity (OB) progression and brain structural changes. T1-weighted magnetic resonance images were acquired from 258 participants with overweight (OW) or OB and 74 participants with normal weight. Participants with OW or OB were divided into four groups according to BMI grades. Two-sample t tests compared disparities between the four subgroups and the participants with normal weight. We used causal structural covariance networks to examine the progressive impact of OB on brain structure. With increasing BMI values, reductions in gray matter volume originated in the left caudate nucleus, medial orbitofrontal cortex, and left insula and expanded to the right hippocampus and left lateral orbitofrontal cortex and then to the right parahippocampal gyrus, left precuneus, and left dorsolateral prefrontal cortex (p < 0.05, false discovery rate corrected). The left caudate nucleus and medial orbitofrontal cortex are the primary hubs of the directional network, exhibiting positive causality to the right hippocampus and left dorsolateral prefrontal cortex. Moreover, the right hippocampus is identified as an important transition hub. These findings suggest that changes in gray matter volume in individuals with OB may originate from reward/motivation processing regions, subsequently progressing to inhibitory control/learning memory regions, providing a new reference direction for clinical intervention and treatment of OB.

Causal Network Representations in Factor Investing

ABSTRACTThis paper explores the application of causal discovery algorithms to factor investing, addressing recent criticisms of correlation‐based models. We create novel causal network representations of the S&amp;P 500 universe and apply them to three investment scenarios. Our findings suggest that causal approaches can complement traditional methods in areas such as stock peer group identification, factor construction, and market timing. While causal networks offer new insights and sometimes outperform correlation‐based methods in terms of risk‐adjusted returns, they do not consistently surpass traditional approaches. The causal method though shows promise in identifying unique market relationships and potential hedging opportunities. However, its practical implementation presents challenges due to computational complexity and interpretation difficulties. Our study demonstrates the potential value of causal discovery in factor investing, while also identifying areas for further research and refinement.

Improving tunnel fire safety: a Bayesian network model for risk analysis and a novel risk acceptability range for risk evaluation

ABSTRACT Event Tree Analysis (ETA) and FN curves are widely used for fire risk analysis and evaluation. However, both have been criticised for their drawbacks. This study addresses these gaps by proposing a Bayesian Network (BN) risk analysis model using fire incident data in Australia, which has never been explored before. The BN model is developed using a score-based approach and validated using four indicators. The innovation of this paper is the introduction of a novel risk acceptability range as a risk evaluation method by integrating BN and Causal Network Topology Analysis (CaNeTA). CaNeTA identifies critical variables through network robustness analysis, which are then used to test sensitivity against current average performance. The results highlight the importance of a rapid and efficient response to fire incidents in road tunnels in Australia. The risk acceptability range represents the current average performance of road tunnels based on actual incidents in road tunnels in Australia. This range is an alternative risk evaluation tool in the absence of specific acceptance criteria. It provides an alternative perspective on risk acceptability in the absence of fatality data. In addition, the process of converting a BN model into CaNeTA provides valuable insights into causal network analysis.