Autocatalysis is an important feature of metabolic networks, contributing crucially to the self-maintenance of organisms. Autocatalytic subsystems of chemical reaction networks (CRNs) are characterized in terms of algebraic conditions on submatrices of the stoichiometric matrix S. Here, we derive sufficient conditions for subgraphs supporting irreducible autocatalytic systems in the bipartite König representation of the CRN. On this basis, we develop an efficient algorithm to enumerate autocatalytic subnetworks and, as a special case, autocatalytic cores, i.e., minimal autocatalytic subnetworks, in full-size metabolic networks. The same algorithmic approach can also be used to determine autocatalytic cores only. As a showcase application, we provide a complete analysis of autocatalysis in the core metabolism of E. coli and enumerate irreducible autocatalytic subsystems of limited size in full-fledged metabolic networks of E. coli, human erythrocytes, and Methanosarcina barkeri (Archaea). The mathematical and algorithmic results are accompanied by software enabling the routine analysis of autocatalysis in large CRNs.
Extreme events such as earthquakes, floods, and power blackouts often display burst phenomena where multiple extreme events occur in quick succession or in bunches. We show that the network structure plays an important role in bunching of extreme events. We use a model of independent random walkers on a complex network. We find that independent walkers on a network with two clusters connected sparsely show oscillatory behavior between the two clusters. A small cluster sparsely connected with the rest of the network shows correlations and bunching among extreme events. The bunching and correlations emerge naturally in our system though the walkers are independent. Such correlations and bunching are not observed in the large cluster. Thus, these correlations are driven by the network structure. We use several characterization techniques, namely, the recurrence time distribution, autocorrelation function, bursty trains, burstiness parameter, and memory coefficient to quantify the bunching and correlations of extreme events.
Peri-operative medicine is a critical component of contemporary healthcare delivery. Despite significant advancements, peri-operative complications remain a relevant concern. Obtaining reliable risk estimates, identifying potential causes, and studying new interventions, revised policies or implementation of best practices to prevent complications, requires data from a large number of participants. Electronic Patient Record systems offer the opportunity to unlock these data, but the limited standardisation of databases and sharing frameworks available across Europe limit the effective use of the available data. We propose creating a European peri-operative shared data registry with continuous data collection, integrating clinical, bedside monitoring and outcome data in a collaborative network. Such network would facilitate outcomes research, could serve as a platform to optimise clinical practices by fostering quality improvement through benchmarking of care delivered by departments or individual physicians, and could be used to evaluate policy changes. This ESAIC initiative aligns well with the development of the European Health Data Space. This article provides examples of contemporary clinical research and practice evaluation questions to illustrate the need for a European collaborative data-sharing network, highlights inspiring examples of existing data-sharing initiatives and describes a road map to establish such network.
The societal impact of rumor spreading is becoming increasingly severe; yet, current research remains relatively one-sided, typically focusing on either rumor propagation or rumor control while neglecting the confrontational and dynamically evolving relationship between them. To address this gap, we propose a novel confrontation framework for rumor modeling. We extend the classical susceptible-infected-recovered-susceptible (SIRS) model into an ignorant-spreader-stifler-vigilant-ignorant (SIRQS) framework by introducing a vigilant state and a confrontation mechanism, thereby capturing subtle differences in individual states during rumor propagation and in their confrontational behavior toward supervisors. At the same time, supervisors patrol the network through random walks guided by node propagation importance, enabling targeted monitoring of rumor spreaders and individuals with a high risk of spreading rumors. Using a microscopic Markov chain approach, we further characterize heterogeneous node behavior and individual differences, and couple the propagation and supervision processes to model node-state transition patterns. We conduct simulations on networks with three different sizes, various topologies, and a real-world network. The results show that the supervision subject, the preference effects associated with the number of supervisors, and the confrontation mechanism are key factors in supervision, and largely determine the effectiveness of rumor propagation control in the simulations, reflecting the substantial influence of these three mechanisms in real-world spreading scenarios. Finally, through multiple evaluation indicators, we provide references for determining the optimal number of supervisors.
Epidemic outbreaks threaten global health and stability, creating an urgent need for effective resource allocation strategies. Existing studies often neglect dynamic regional risk adjustments and resource coordination based on cluster structures. To address this, this paper proposes an adaptive resource allocation mechanism based on the patch cluster structures and develops a coupled dynamics model that integrates resource flow with epidemic spreading in a metapopulation network. The model employs a migration-interaction-return process to characterize both inter-patch migration and intra-patch epidemic spreading. Furthermore, an adaptive resource allocation mechanism is designed, which dynamically adjusts both inter-patch donation strategies and intra-cluster allocation schemes according to evolving, patch-specific risk levels, thereby realizing dynamic optimization of resource distribution. Using the micro Markov chain approach, we derive epidemic evolution equations and calculate infection thresholds. Numerical simulations validate the model and examine key parameter impacts. The results show that increasing patch cluster numbers, enhancing inter-cluster connectivity, and improving cluster efficiency-especially in networks with abundant triangular structures-effectively raise the epidemic threshold and reduce infection scale. Compared to traditional models, adaptive resource allocation models can utilize resources more efficiently, thereby decreasing the infection scale. Higher donation/utilization rates mitigate global spread, while targeted assistance from high-risk to low-risk patches lowers overall prevalence. This study provides a theoretical framework for dynamic group resource optimization in heterogeneous risk environments, offering valuable insights for epidemic prevention and control.
Walking impairments post-stroke are common but studying their neural basis with fMRI is difficult due to motion constraints. This study used post-task resting-state residual activity (PTRRA) to investigate walking-related neural activity without in-scanner movement. Ten patients with subcortical ischemic stroke and ten matched healthy controls underwent two resting-state fMRI sessions: one immediately after overground walking, and another after a 20-minute delay. Clinical assessments included Fugl-Meyer Assessment and Timed Up-and-Go Test (TUGT). Analyses included ALFF, functional connectivity (FC), and graph theory. Significant group × time interactions showed that stroke patients had increased ALFF in the contralesional primary somatosensory cortex (S1) during the delayed session. This increase inversely correlated with TUGT scores (r = -0.677, p = 0.032). Enhanced FC was found between contralesional S1 and ipsilesional sensorimotor regions. ALFF changes and bilateral S1 FC were significantly associated with mobility outcomes. Graph theory results were atlas-dependent. Increased contralesional S1 activation and its enhanced connectivity with bilateral motor areas may reflect adaptive changes supporting post-stroke balance. The PTRRA method proved feasible for capturing post-walk neural activity and may be useful for assessing large-scale network dynamics after stroke.
To compare the effects of different exercise modalities on depressive and anxiety symptoms in patients with cancer and to explore the exercise dose-response relationship to identify the optimal dose. Randomized controlled trials published from database inception to January 2026 were searched in PubMed, Web of Science, Embase, and the Cochrane Library. Risk of bias was assessed using the Cochrane Risk of Bias tool. Stata 17.0 and R 4.4.3 were used for data transformation and statistical analyses, including comparisons of exercise modalities and dose-response evaluation for depressive and anxiety outcomes. Sixty-seven randomized controlled trials involving 5778 patients with cancer were included. Network meta-analysis showed that mind-body exercise (MBE), combined aerobic and resistance exercise (COM), and aerobic exercise (AE) significantly improved depressive symptoms, whereas resistance training (RT) did not. For anxiety, significant improvements were observed with MBE and AE, with MBE showing the greatest benefit for both outcomes. Dose-response analysis showed a nonlinear U-shaped association between total exercise dose and both depression and anxiety, with optimal doses of 770 and 700 MET-min/week, respectively. Different exercise modalities vary in their effects on depressive and anxiety symptoms in patients with cancer, with MBE showing the greatest overall benefit. The U-shaped dose-response relationship suggests that an appropriate exercise dose may help optimize improvements in depressive and anxiety symptoms and provide a basis for more precise exercise prescriptions based on exercise modality and dose.
We propose the Optimal-Transport Gated Echo-State Network (OT-ESN), a two-timescale reservoir that replaces ad hoc inter-module couplings with a principled, mass-conserving transport mechanism on a cortical-sheet geometry. At each step, a slow, exogenous controller computes an entropically regularized optimal-transport plan Π between the previous distribution of column activity (source) and an input-derived "intent" over columns (target), using a geometric cost that encodes anatomical or functional proximity. The resulting plan-doubly stochastic up to prescribed marginals-acts as a bounded, geometry-aware mixer that gates inter-column blocks of the reservoir at the next fast update. This one-step delay ensures that Π is absent from the time-t Jacobian, so with a 1-Lipschitz nonlinearity and fixed leak, the echo-state property collapses to a single spectral-norm inequality on pre-scaled intra- and inter-column operators, yielding a uniform contraction certificate. OT-ESN, thus, achieves interpretable, neuromodulation-like routing of assembly activity while preserving the simplicity of readout-only training. Computationally, Sinkhorn iterations on a J×J kernel provide efficient, smooth control, with the regularization parameter spanning diffuse (diffusion-like) to sharp (path-like) transports without jeopardizing stability. Ergo, via optimal transport, OT-ESN enables long, structured memory and geometry-respecting information flow in a provably stable recurrent substrate.
The pKa values of titratable residues are fundamental parameters that shape protein behavior in a given solution environment. They govern electrostatic interactions and thereby influence protein folding, conformational dynamics, and molecular recognition. Existing pKa-prediction algorithms typically rely on detailed all-atom structures. However, although coarse-grained models are widely used to simulate protein structure and dynamics, estimating pKa values within these frameworks remains challenging because ionizable groups are often not explicitly represented. In this work, we introduce DeepCGpKa, a deep-learning-based pKa predictor designed for coarse-grained protein structures. Benchmark tests show that DeepCGpKa attains accuracy comparable to state-of-the-art methods based on all-atom structures. It also retains robust predictive performance for partially unfolded protein structures. When coupled with coarse-grained molecular simulations, DeepCGpKa successfully captures the pH dependence of protein conformational changes. Integrating data-driven pKa prediction with physics-based molecular simulations provides a practical route to improve the treatment of electrostatic interactions at the coarse-grained level, which is a critical issue for most currently available coarse-grained biomolecular models.
The prevalence of metabolic dysfunction-associated steatotic liver disease (MASLD) continues to rise, underscoring the need for tools to stratify individual risk of disease progression. We evaluated whether logistic regression models augmented by deep learning-based predictions (DLPs) can improve the B-mode ultrasound-based identification of at-risk MASLD, defined as patients with increased fibrosis risk. We retrospectively analyzed 205 patients with a total of 636 ultrasound images. We developed a model that reproduces the LSM-based dichotomous fibrosis risk classification using clinical parameters and ultrasound image-derived deep learning pipelines. Patients were classified by same-day liver stiffness measurement (LSM) (<8 kPa: low fibrosis risk; ≥8 kPa: increased fibrosis risk). We assessed the incremental value of DLPs when added to the parameters sex, age, BMI, diabetes mellitus type 2 status and the fibrosis-4 score (FIB-4) based on accuracy, AUROC, and related statistics. The logistic regression model combining the clinical parameters and the DLPs achieved acceptable performance with an AUROC of 0.73 and a test accuracy of 68%. The same model without DLPs showed an AUROC of 0.72 and a test accuracy of 61%. Including FIB-4 improved performance further (AUROC 0.92, accuracy 88%). Models based solely on image data demonstrated limited diagnostic performance. B-mode ultrasound provides a weak fibrosis-related signal, yielding limited diagnostic performance. Meaningful discrimination required the incorporation of clinical parameters, with FIB-4 offering the greatest improvement among the parameters assessed. Deep learning predictions added only modest incremental value. Prospective validation is needed to clarify clinical utility.
Meat-borne Staphylococcus aureus (S. aureus) remains a leading global foodborne pathogen harbouring novel enterotoxin genes (NEGs) encoding superantigenic toxins with conditionally enhanced pathogenicity, representing a critical food safety hazard. This review characterizes NEG features, pathogenic mechanisms, multi-layered regulatory networks, and identifies key research challenges. We systematically searched PubMed, Embase and the Web of Science Core Collection to synthesize molecular and multi-omics evidence (ChIP-seq, RNA-seq, CRISPR-Cas9) characterizing NEG classification, pathogenic mechanisms, multi-level regulatory networks (transcriptional, post-translational, environmental), and geographical distribution patterns in meat-derived S. aureus. NEGs comprise superantigenic and tissue-targeting subgroups mediating pathogenicity via cytotoxicity, intestinal microenvironment disruption, and immune evasion. Regulation involves a complex network of Agr/σB/SarA/Rot-mediated transcriptional control, phosphorylation/lactylation modifications and environmental sensing, exhibiting marked geographical divergence. Current limitations include technical resolution constraints, insufficient physiological model fidelity and incomplete regulatory crosstalk elucidation. Future research should prioritize transcription factor interaction mechanisms, growth-toxin correlation prediction models and multi-omics-based network decipherment. This review provides a foundational framework for NEG research to inform food safety risk assessment and targeted contamination control strategies.
This exploratory study investigated alterations in cerebral metabolism and metabolic connectivity using 18F-fluorodeoxyglucose positron emission tomography (F-18 FDG-PET) in patients with multiple myeloma with neurotoxicity after B-cell maturation antigen (BCMA)-directed chimeric antigen receptor (CAR) T-cell therapy. The retrospective study included 20 BCMA CAR T-cell therapy recipients who underwent brain F-18 FDG-PET imaging (15 with baseline PET scans). Brain metabolism was quantified using standardized uptake value ratios (SUVRs), with the cerebellum as the reference region. Metabolic connectivity was assessed using correlation matrices and network topology metrics. Metabolic patterns were age-adjusted and compared between patients with and without neurotoxicity. Neurotoxicity occurred in 7 patients. SUVR analysis involving those with neurotoxicity demonstrated significantly reduced post-treatment FDG uptake in the left precentral gyrus and striatum. Greater declines were observed in FDG uptake within the left inferior occipital cortex and striatum in patients with neurotoxicity than in those without. Metabolic connectivity analysis identified 508 significantly altered regional pairs, primarily with reduced frontotemporal correlations. Hub node analysis demonstrated a redistribution of network centrality from higher-order cortical regions to temporal and occipital areas, with reduced centrality in the insular and limbic regions among patients with neurotoxicity. Patients with neurotoxicity after BCMA CAR T-cell therapy demonstrated distinct metabolic-level and network-level differences on F-18 FDG-PET, highlighting the potential of brain PET imaging to elucidate the underlying mechanisms and warranting further investigation in larger cohorts.
Donation after circulatory death (DCD) is an emerging heart transplantation (HT) strategy with improved waitlist and comparable post-transplant outcomes to donation after brain death (DBD) in adults. Pediatric DCD-HT is underutilized but gaining broader adoption. We assessed the impact of DCD listing on waitlist outcomes and graft survival in pediatric HT. We queried the OPTN (Organ Procurement and Transplantation Network)/UNOS (United Network for Organ Sharing) database for all pediatric primary isolated HT candidates between January 2021 and June 2024. Waitlist outcomes were compared by final listing type to account for crossovers. Offer dynamics were assessed by average interval between offers stratified by OPTN/UNOS status. Multivariable regression modeled offer frequency as a function of DCD listing. We compared 30-day survival between recipients of DCD and DBD organs. Among 2486 candidates, 86 were initially listed eligible for DCD and 2400 for DBD. DCD candidates had higher clinical acuity than DBD candidates. DCD listing was associated with significantly shorter intervals between offers across all statuses and increased offer rate by 134% (95% CI: 92%-186%). Waitlist outcomes did not differ significantly by final listing type. There was no difference in 30-day survival between DCD and DBD recipients. DCD listing in pediatric HT is associated with a shorter interval between offers and more frequent offers. Waitlist survival was similar between groups despite DCD candidates being sicker at listing. There was no difference in 30-day survival between DCD and DBD recipients. These findings suggest that broader adoption of pediatric DCD-HT can expand access to donor hearts without compromising early post-transplant outcomes.
Drawing on Miranda Fricker's "hermeneutical marginalization" and Martin Meeker's "sexual communication network," this article retraces how the adoption and therefore the existence of queer identities are predicated on the availability of alternative models of understanding sexuality. "Loci of increased human connection," such as cities, media, and the Internet, are recognized as disseminators of hermeneutical resources and catalysts for "sexual communication networks." By following this throughline, this article provides a framework for the social ontology of sexual orientation that is intended to be usable across history. The author renegotiates the debate about the applicability of sexual orientation lingo to queer pre-modern history by iterating on William Wilkerson's emerging fusion theory of sexual identity. Sexual orientation is a self-interpretation of desires, which emerge, or remain hermeneutically marginalized, under the available social models that are specific to each society (and time). Historiography, in this way, can realistically enquire about sexual preference in its search for "gay history"; particular focus must be put, however, on "loci of increased human connection," because they are the most likely sites where hermeneutically underserved desires might emerge explicitly as alternative social categories. These alternative categories might have lived only in limited contexts before the invention of modern communication technologies.
This work proposes a general framework for Neural Delay Differential Equations (NDDEs), a class of continuous-depth neural networks, namely, the Generalized NDDEs (GNDDEs), incorporating various types of delays beyond the previously considered constant delays, including time-dependent, state-dependent, and time-state-dependent delays. Furthermore, we employ a simulation-free training strategy for the vector field, allowing the system reconstruction directly from the irregularly sampled time series without the prior model knowledge. Specifically, we perform the regression between the preprocessed target and parameterized vector fields, bypassing the need to numerically solve the differential equations as required in conventional time-series regression. Additionally, GNDDEs enable adaptive, model-free identification of the delay functions, along with the model-based identification of the system parameters. The experimental results demonstrate that the proposed framework exhibits the notable effectiveness and computational efficiency across a variety of delay differential equation problems, further broadening the applicability of continuous-depth neural networks in the delay system modeling.
ObjectiveTo systematically review literature on the use of artificial intelligence (AI) and machine learning (ML) models for detecting velopharyngeal dysfunction (VPD) in patients with cleft palate.DesignSystematic review conducted in accordance with PRISMA guidelines (PROSPERO CRD420251034524).SettingStudies published were identified through EMBASE, ProQuest, Google Scholar, and PubMed.ParticipantsA total of 3967 participants contributed 92,323 training samples. Internal validation included 2331 controls and 2449 VPD cases, generating 81,143 validation samples. Ages ranged from 1 to 93 years.InterventionsML models were trained on speech features such as mel frequency cepstral coefficients (MFCCs) and constant Q cepstral coefficients (CQCCs) to classify or validate VPD-related speech outcomes.Main Outcome Measure(s)Reported performance metrics included accuracy, precision, recall, F1-score, sensitivity, specificity, and Pearson correlation coefficient (PCC). External validation was assessed when reported.ResultsOf 455 screened articles, 34 met the inclusion criteria. Support vector machines were the most commonly used models (16/34, 47.1%), followed by convolutional neural networks (6/34, 17.6%) and deep neural networks (2/34, 5.9%). Across studies reporting performance metrics, midpoint estimates yielded a mean accuracy of 82.9%, precision of 86.7%, F1-score of 0.88, sensitivity of 80.5%, specificity of 82.2%, and PCC of 0.58. Only 3 studies (3/34, 8.8%) performed external validation.ConclusionsAI/ML models demonstrate promise for VPD detection with encouraging performance. Inconsistent reporting, reliance on engineered features, and limited external validation restrict generalizability. No clinically deployable model has yet been achieved.
Effective pest management requires accurate and continuous monitoring. This monitoring helps assess population dynamics and guides the development of integrated pest management strategies. Traps used to capture insects are an alternative applied to various crops. However, the identification and manual counting of specimens are time-consuming, require taxonomic knowledge, and depend on the expertise of specialists. Automation could reduce costs, increase accuracy, and enable scalable analyses. Current computer vision and artificial intelligence techniques can quickly and accurately identify objects in digital images. This study presents a systematic review of literature retrieved from multidisciplinary and specialized databases (Scopus, ACM, Web of Science, IET, DBLP, Springer, and ScienceDirect), focusing on the intersections of agriculture, ecology, and computer science. We found 284 studies published between 2020 and 2025. Among them, 57 fulfilled the eligibility criteria, considering applied computing solutions for insect identification and counting using digital images of specimens collected via traps or photographed in situ on plants, in both field and laboratory settings. The findings highlight the use of electronic traps for real-time data collection and improvements in convolutional neural networks, with visual transformers and attention mechanisms for multi-species and fine-grained recognition. They also indicate opportunities to leverage microscopy resources, overcome limitations in the large-scale deployment and integration of electronic trap networks, and integrate real-time monitoring data with forecasting models using weather predictions to promote early warning systems for integrated pest management.
Vision-threatening ocular diseases are impacted by aging-associated molecular changes, including mitochondrial dysfunction, cellular senescence, and chronic inflammation. Anti-VEGF therapies targeting VEGF-A/VEGFR2 signaling remain the frontline standard of care, but many patients exhibit suboptimal or nondurable responses, often due to compensatory and/or compromised antiangiogenic and anti-inflammatory pathways. We aimed to elucidate shared mechanisms underlying treatment failure and disease progression. We applied an integrative systems biology framework that combined multiomics datasets, network-based machine learning, and disease-specific pathway mapping. A comprehensive literature review of conditions, including diabetic retinopathy, age-related macular degeneration, retinitis pigmentosa, glaucoma, and aging, identified 14 core genes consistently associated with angiogenesis, inflammation, and immune signaling. Multialgorithm centrality and enrichment analyses reconstructed disease-specific interaction networks, revealing consensus mechanistic axes. Integration of cell-type-specific single-cell RNA sequencing data from AMD-RPE clusters identified cluster-specific gene hubs and vertical signaling axes, leading to VEGF blockade failure. EGFR, HSP90AA1, SIRT1, and STAT3 emerged as central resistance hubs linking angiogenesis and inflammatory processes. Pathway enrichment analyses revealed 21 conserved core signaling cascades, grouped into six functional categories, with AGE-RAGE, PI3K-Akt, HIF-1, MAPK, and chemokine pathways playing central roles. A MiRGD-based peptide nanocomplex delivering htsFLT01 achieved efficient RPE transfection and controlled gene activation under basal conditions. This systems-level framework clarifies mechanisms of VEGF blockade resistance and provides a rational basis for next-generation, combinatorial therapeutic strategies requiring validation in disease-relevant models.
Traditional drug discovery is a resource-intensive process with high attrition rates and the huge difficulty of working with a chemical space that is thought to include [Formula: see text] molecules. Even though computational chemistry has come a long way, traditional generative models still use string-based representations like SMILES, which have trouble capturing intricate three-dimensional spatial interactions and often make structures that aren't real. Moreover, current reinforcement learning methodologies frequently do not achieve an equilibrium between molecular diversity and high-affinity biological activity. To overcome these constraints, this research introduces an innovative integrated framework that merges Geometric Multi-Discrete Soft Actor-Critic (Geom-SAC) and Multi-stage Variational Autoencoders (MS-VAE) to improve de novo molecular creation and activity optimisation. The main new idea is the combination of geometric deep learning, which enforces physical atomic restrictions, and a hierarchical VAE architecture, which organises the latent space into manageable structural steps from scaffold formation to functional group optimisation. We also use a Non-Covalent Interaction-Aware (NCIA) graph neural network in our method to improve protein-ligand affinity predictions by simulating complex intermolecular forces. Experimental results on benchmark datasets, such as ZINC250k and PDBbind, show that the proposed framework improves binding affinity scores by 15% and the Valid-Unique-Novel (VUN) molecule ratio by 20% compared to the best existing methods. Also, adding a security layer based on blockchain technology makes sure that data is secure and can be tracked. This all-encompassing method provides a strong, highly accurate answer for next-generation AI-driven pharmacology. It greatly narrows the gap between computational design and experimental validation.
Symbioses between mites and beetles are ubiquitous, but the relationships among mites and leaf-feeding, free-living beetle species have not been well studied. To clarify the relationships between a phoretic mite, Coleolaelaps longisetatus, and leaf-feeding, free-living Polyphylla beetle species in Japan, we determined phoretic rates of mites and clarified the patterns of genetic diversification of the mite in relation to the host beetle species. Totals of 252 P. albolineata, 31 P. laticollis, and 44 P. schoenfeldti were collected from 55, 14, and 10 sites, respectively. Coleolaelaps longisetatus was found on 224 P. albolineata (89%), 28 P. laticollis (90%), and 42 P. schoenfeldti (95%). No geographic pattern was observed in phoretic rates of mites. Phylogenetic analyses indicated that the C. longisetatus collected from each host beetle species formed monophyletic groups, suggesting that the diversification process of the host beetles affected the divergence of the phoretic mites. Haplotype network analyses showed that there was a roughly geographic pattern in genetic diversity within the mite clades, which may reflect the dispersal abilities of the host beetles.