Abstract: Density Peak Clustering (DPC) is a widely studied unsupervised clustering method that identifies cluster centers using local density and relative separation without iterative prototype updates. Its simple and intuitive formulation has motivated extensive research, but its performance remains sensitive to data representation, distance measures, neighborhood construction, density estimation, center selection, assignment strategies, and data scale. To provide an updated and structured view of the field, this paper presents a systematic mapping review of DPC research published from 2014 to 8 August 2026. The survey first establishes the fundamental concepts, original DPC procedure, methodological evolution, and current methodological landscape. It then examines major application areas, including healthcare, cybersecurity, industrial monitoring, remote sensing and image analysis, transportation, networks, text analysis, multivariate time series, and scientific data. Subsequently, the methodological advances enabling modern DPC are organized into eight families: adaptive neighborhood and density estimation; graph, manifold, and multi-peak modeling; automatic center and cluster-number determination; assignment, merging, and uncertainty handling; representation learning; privacy-preserving, federated, and incremental computation; scalable execution; and auxiliary strategies such as sampling, grids, optimization, and limited supervision. Each methodological family is related to the limitation it addresses and the assumptions or computational costs it introduces. Finally, the survey synthesizes key lessons, research potential, persistent challenges, and future directions toward robust, interpretable, fairness-aware, privacy-preserving, reproducible, resource-efficient, uncertainty-aware, and continually adaptive DPC systems.

Abstract: Background: Long COVID (LC) affects millions worldwide, causing diverse symptoms, from fatigue, dyspnea, and cough to multiorgan complications. Immune activation, endothelial dysfunction, hypoxia, and metabolic dysregulation are implicated, although mechanisms remain unclear. This study aimed to identify a biomarker panel linking endothelial dysfunction, hypoxia, and metabolic shifting to diverse LC symptoms. Methods: Plasma biomarkers were compared between 30 individuals with LC and 30 controls. Glucotoxic methylglyoxal, glyoxalase-1 and glutathione, indicators of methylglyoxal detoxification, soluble VAP-1 (sVAP-1, SSAO), TNF-α, and IL-1β were assessed. STRING and molecular docking explored protein associations and potential MG–Glo-1 interactions. Results: Compared with controls, individuals with LC exhibited approximately 200% higher MG, 20% lower glyoxalase-1, 45% lower glutathione, and 50% higher SSAO activity. TNF-α and IL-1 increased 165% and 145%, respectively. Symptom numbers correlated with methylglyoxal (r = 0.65) and inversely with Glo-1 (r = −0.61) and GSH (r = −0.47), but not SSAO, TNF-α, or IL-1β. Methylglyoxal was inversely correlated with glyoxalase-1 (r = −0.59) and GSH (r = −0.67), but not SSAO, TNF-α, or IL-1β. STRING analysis predicted associations involving glyoxalase-1, VAP-1, and other proteins, while docking suggested MG interaction with glyoxalase-1(−4.82 kcal/mol). Conclusion: MG/Glo-1/glutathione represents a novel biomarker panel for diverse LC symptoms.

Abstract: Objective: To map how lactate has been measured, interpreted and validated as a graft viability or outcome marker across donor, perfusate and recipient phases and organs. Background: Lactate is measured at every step of solid organ transplantation, but its meaning differs by phase, preservation method and organ. No review has mapped this evidence across the transplant continuum. Methods: Scoping review following JBI methodology and PRISMA-ScR. PubMed/MEDLINE, Embase, Web of Science and CENTRAL were searched from 2000, supplemented by journal-targeted searches, citation chasing and ClinicalTrials.gov. Human clinical and ex vivo studies reporting lactate against graft viability or outcome were charted and mapped by organ and phase. Results: 159 human studies were included. Liver transplantation accounted for 108, mainly in the recipient phase and machine perfusion; heart contributed 28, lung 12, kidney 7 and pancreas 1, and no intestine study was eligible. Proposed liver perfusion thresholds ranged from below 1.7 to 4.1 mmol/L. In the largest cohort, lactate clearance did not predict graft survival (AUC 0.50). Day-1 lactate contributed to a validated primary nonfunction model (AUROC 0.83). Confounders included stored red cells in circuits, absence of a clearing organ, donor type, measurement timing and drugs. Conclusions: Lactate is a useful but organ- and context-dependent signal, strongest as a recipient-phase marker after liver transplantation and weakest as a stand-alone ex situ viability criterion outside the liver. Phase-specific thresholds, standardized reporting and multi-marker validation are needed before lactate is used to discard organs.

Abstract: Inflammatory bowel disease (IBD) is increasingly recognized as a disorder in which extracellular matrix (ECM) remodeling is not merely a downstream consequence of mucosal inflammation but a potential significant contributor in pathogenesis. The intestinal ECM regulates epithelial barrier integrity, immune cell trafficking and the resolution of inflammation. Evidence derived predominantly from murine models demonstrates that structural, compositional and mechanical alterations may emerge before clinical symptoms. This pre-symptomatic state is characterized by subclinical immune infiltration and elevated matrix remodeling enzyme activity within the intestinal epithelium. During active disease, remodeling intensifies, with increased matrix metalloproteinase activity generating ECM fragments that reflect endoscopic severity and may discriminate flares from remission. This review synthesizes the evidence on ECM remodeling during the pre-symptomatic and flare phases, emphasizing the ECM as an active driver of early inflammation. Recognizing this early window opens opportunities for diagnostic and therapeutic strategies beyond the current focus on established mucosal inflammation.

Abstract: Background. Knowledge-graph models can rank existing drugs for a disease, but their benchmarks mostly reward rediscovering known treatments, and their scores rarely come with evidence a researcher can check. We built OREVID, an open tool that shows a ranked candidate only when an evidence line supports it and reports how often each line was right on held-out data, and compared it with TxGNN on identical data. Methods. The graph combines PrimeKG, Open Targets, STRING and PubTator3 (157,822 nodes, 9.46 million edges). For each disease, 22 degree-weighted metapath features and two gene-level columns (whether the drug moves its targets in the direction human genetics asks, from protein and expression QTLs that colocalise with the disease's genome-wide association studies, and whether it acts on a gene Orphanet lists for the disease) feed a LightGBM LambdaRank model trained with masked neighbourhoods. A candidate is shown only if it is indicated for an ontology-neighbour disease or for a disease sharing at least 3 associated genes (hypergeometric p < 10⁻⁵ over the genes with at least one associated disease). The acceptance criteria for these rules were set in advance; the rules were revised on the hidden-treatment benchmarks and then measured on an external set of approvals not used to choose them. TxGNN was retrained with its official code on the same PrimeKG splits, and all tasks and predictions are released as OREVID-bench. Results. Zero-shot Recall@100 (mean ± SD over five splits) was 0.812 ± 0.038 for OREVID, 0.412 ± 0.042 for TxGNN and 0.358 ± 0.040 for a popularity baseline, with OREVID ahead on every split (paired difference with TxGNN 0.400, 95% CI 0.354–0.444, resampling diseases). On nine disease areas held out it was 0.627 against 0.237, ahead in all nine, although TxGNN had the higher AUPRC in four. Training OREVID without validation labels and excluding diseases where TxGNN outputs constant scores gave 0.815 against 0.421. On 3,718 approvals absent from PrimeKG, Recall@100 was 0.355 against 0.121 (0.366 ± 0.011 against 0.126 ± 0.004 over five training splits), and 0.175 against 0.085 for the 2,719 approvals not obvious from the disease ontology; a logistic regression without the drug-popularity feature and a model without any Open Targets feature (0.348) performed similarly. Of the candidates the website's rules select in this external exam, 12.6% (95% CI 9.8–15.5%, resampling diseases) of direct extensions and 4.4% (3.2–5.8%) of shared-genes candidates were among the unseen approvals, against about 0.07% by chance; random gene sets of the same size selected candidates that were right as often, so the shared genes are not evidence of a shared mechanism, although they contained a protein of the curated mechanism (DrugMechDB) more often than random genes of the same disease (25% against 17%). A drug acting on a disease gene, with no other evidence, was right 1.8–3.4% of the time, no better than candidates with no evidence line at all. With the graph and the model rebuilt as known in 2015 and every drug dated by its first approval, the uses that later reached phase 3 or approval had a Recall@100 of 0.283 against 0.181 for popularity (paired difference 0.102, 95% CI 0.078–0.126) and 0.117 for TxGNN trained on the same graph (difference 0.167, 0.142–0.192), in pre-registered tests. On labels absent from the graph (CIViC oncology evidence, indications only DrugCentral or MED-RT record, and European orphan designations) OREVID was ahead of popularity by 0.32, 0.27 and 0.16, but the candidates the website shows did not meet trial end points more often than ranked drugs it hides. Conclusions. Reporting the hit rate of each evidence line makes repurposing predictions checkable and shows how low their precision is on approved uses missing from the training graph. Most of the signal comes from the ranking and from what already works in related diseases, not from the specific genes they share. OREVID is a hypothesis generator, not a clinical tool; code, data and benchmark are public (https://orevid.org, doi:10.5281/zenodo.23323226).

Abstract: AI has become a core component of consumer-facing marketing activities ranging from personalised product recommendations, conversational shopping assistants, targeted advertising, dynamic content and predictive offers. These uses of AI can affect how consumers search, consider and choose between different products. Yet, the extent to which these AI applications will affect purchase behaviour will not only depend on their technological capabilities. For AI-based marketing to influence consumer choice, consumers need to view the marketing communications as relevant, trustworthy and useful (see Ribeiro et al., 2025, for similar findings). The study investigates the effect of AI-enabled personalisation on purchase intention in consumers with specific focus on consumer trust and perceived value. The study will adopt a quantitative, cross-sectional approach. The population of the study is consumers who have encountered the use of AI-enabled recommendations or personalised marketing on digital channels. A structured questionnaire of five-point Likert scale will be used and a sample size of 250 respondents was used as an example for the statistical analysis alone because the primary data was not provided. Descriptive statistic, Pearson correlation and multiple regressions are proposed to be used.The results shown on this page, the following positive effects are observed between the variables: AI personalisation -> trust -> perceived value -> purchase intention. This is based on the example regression analysis results showing perceived value and trust have a great impact on purchase intention, AI personalisation has a positive direct relationship as well. However, bear in mind that these results are only an example, and should never be reported as actual research.

Abstract: Background/Objectives: The LUNAR trial demonstrated the benefit of Tumor Treating Fields (TTFields) therapy in patients with metastatic NSCLC (mNSCLC) after platinum-based chemotherapy. However, real-world outcomes in patients treated with contemporary systemic regimens have yet to be empirically demonstrated. Methods: In this retrospective single-institution study, patients with mNSCLC treated with TTFields (150 kHz) after platinum-based chemotherapy between 2018 and 2026 were included. Progression-free survival (PFS) and overall survival (OS) were estimated from TTFields therapy initiation using the Kaplan–Meier method. Results: Sixteen patients were included. The median age was 73 years, 63% were male, 69% had adenocarcinoma histology, and 69% were PD-L1 positive. At TTFields therapy initiation, 94% had ECOG performance status ≤ 2, 19% had brain metastases, 31% had liver metastases, and 94% had prior ICI exposure. Median PFS was 11.8 months, and median OS was 19.9 months. OS did not differ significantly between the ICI-based and docetaxel-based groups (not reached vs. 13.5 months; p = 0.60) but PFS was shorter (3.9 months vs. not reached; p = 0.046). Treatment-related dermatitis was limited to grade 1–2 events, and no grade 3 or higher device-related toxicities were observed. Conclusions: TTFields therapy demonstrated encouraging real-world OS outcomes and favorable tolerability in this first report of a contemporary post-platinum NSCLC population, supporting further prospective evaluation of optimal timing and integration with modern systemic therapies.

Abstract: The same object state can produce different responses under different observation-formation conditions. We ask which information must be retained once the identities of the two sides and the response task have been specified independently, and how this requirement changes with the prediction horizon and the scale of description. Constancy of the response on fibers characterizes joint sufficiency and the eliminability of either side. Including identity in the output allows the standard behavioral quotient construction to yield the coarsest closed state representation that preserves the prescribed identities. Nested tasks under a common evolution induce a compatible family of predictive scales. Exact closure, however, does not ensure that observation conditions remain non-eliminable after coarse-graining: the roles of both sides and a response difference must also survive. A model of two coupled chains gives explicit expressions for the required initial state and the fraction of a response difference retained by a mode projection; a nonuniform chain illustrates the failure of closure. Independent calibration and held-out evaluation use synthetic data, and their numerical results agree with the full classical theory of coupled dynamics. The contribution is a unified analysis of identity constraints, predictive closure, and the retention of response differences, supported by reproducible model calculations. No new theory of minimal realization, or new physical effect implied by the dual-axis structure alone, is claimed.

Abstract: Pavement cracking is a common form of distress in asphalt pavements, and crack detection is essential for pavement maintenance. To improve detection accuracy in complex backgrounds, this study develops an enhanced YOLOv11-based detection algorithm. First, we construct CACrack2025, a pavement crack dataset that provides samples for model training and evaluation. Second, we design CrackFormerV2, a feature extraction module that combines cross-scale feature interaction with contextual information aggregation to improve the representation of slender, low-contrast cracks. To address the imbalance between crack targets and background regions, we incorporate Adaptive Threshold Focal Loss (ATFL), which reweights easy and hard classification samples according to their confidence. Experiments on CACrack2025 show that the proposed method achieves an mAP@50 of 86.0%, 4.6 percentage points higher than the baseline YOLOv11n model. Under the current dataset and evaluation protocol, the method maintains higher detection accuracy than the evaluated baselines. These results support the effectiveness of the proposed components for pavement crack detection.

Abstract: Causal analyses usually treat missing outcome measurements as a nuisance process that can be handled after the physical system has been modeled. Extreme hazards can violate that separation: the latent hazard can destroy the instrument needed to observe it. We study this problem using the 26 August 2026 Nepal Bhote Koshi-Trishuli disaster. Let H denote latent local flood/debris-flow severity and R indicate whether a gauge remains capable of reporting. The central causal structure is H -> R. If an exposure A changes H, and higher H also lowers the probability that R=1, then a complete-case analysis preferentially excludes some of the most severe outcomes and can bias the causal contrast E[H(1)-H(0)]. Direct ten-minute Department of Hydrology and Meteorology records show that Rasuwagadhi, Syabrubesi, and Betrawati stopped reporting after last values below their warning thresholds, while surviving downstream stations subsequently recorded the flood wave. We therefore treat each failed station's last valid stage as a lower bound on its unobserved event peak rather than as the peak itself. Downstream hydrology and Copernicus EMSR927 damage mapping are used as auxiliary constraints, not as replacements for the missing stage. In a separate truth-known Nepal-calibrated simulation with a known ATE of 4.65 standardized peak-anomaly units, complete-case analysis under hazard-dependent censoring underestimates the effect by 25.1%; linear interpolation and MAR-IPCW remain biased by 16.4% and 16.0%, whereas failure-aware estimators are within 1% of the truth. The paper's main contribution is therefore a causal framework for catastrophe-induced informative censoring: the disaster changes both the physical state and the probability that the state remains measurable. The real event supports partial identification—sets of peak values compatible with observed lower bounds and physical constraints, rather than unsupported point reconstructions—and the simulation quantifies the causal bias produced when this observation process is ignored.

Abstract: To improve the winter snow-melting and temperature-regulating performance of asphalt mixtures, a low-temperature phase change material (LPCM) was prepared by melt blending polyethylene glycol 400 (PEG400) with expanded graphite at a mass ratio of 9:1. The prepared LPCM and sustained-release deicing salt were incorporated into SMA-13 asphalt mixtures. Five mixtures were designed, including a control mixture, an LPCM-modified mixture, a salt-modified mixture, and two mixtures containing both additives at different dosages. Differential scanning calorimetry (DSC) and Fourier transform infrared spectroscopy (FTIR) were used to characterize the materials. The combined effects were evaluated through laboratory ice-melting tests, temperature–time measurements, natural snowfall observations, and pavement performance tests. The results showed that the melting and crystallization enthalpies of LPCM were 91.1 and 88.1J/g, respectively. Physical adsorption and mixing were the dominant mechanisms during the preparation and modification processes. Compared with the control mixture, the mixtures containing LPCM reached 0℃ for the first time 6–24min later. The minimum temperature of the high-salt composite mixture increased by 1.7℃, while its average cooling rate decreased by 9.1%. The delayed cooling helps maintain a liquid phase and favorable wetting conditions at the ice–pavement interface. These conditions facilitate the continuous dissolution and migration of chloride salts from the carrier, allowing thermal regulation and freezing-point depression to act in a complementary manner. The high-salt composite mixture achieved a relative ice-melting capacity of 0.34 g/cm² after 3h, representing a 21.4% increase over the mixture containing the same salt dosage alone. During the first 6 h of outdoor observation, its average snow-melting rate, calculated from changes in snow layer thickness, reached 0.079 mm/min. This was 26.7% higher than that of the control mixture, indicating improved ice-melting and snow-removal performance. The low-salt composite mixture exhibited a dynamic stability of 4755times/mm and a flexural tensile strain of 3571με. Its residual Marshall stability and tensile strength ratio (TSR) were 93.0% and 90.2%, respectively. Increasing the salt dosage further enhanced ice-melting and snow-removal performance but reduced rutting resistance and moisture stability. The combined incorporation of LPCM and sustained-release deicing salt provides both cooling mitigation and ice–pavement interface regulation. The low-salt composite mixture achieves a favorable balance between winter functional performance and pavement performance.

Abstract: Control-valve stiction is a persistent source of nonlinear oscillation, degraded regulation, and excessive actuator activity in process-control systems because identical controller-demand values can correspond to different stem positions depending on motion direction and sticking history. This study develops a path-parameterized data-driven framework for reconstructing the resulting multivalued controller-output–valve-position trajectory using four first-order Takagi–Sugeno fuzzy architectures: fuzzy C-means, 16-rule and nine-rule subtractive-clustering models, and an adaptive neuro-fuzzy inference system. A normalized path coordinate converts the multivalued trajectory into synchronized single-valued representations suitable for fuzzy reconstruction. Validation uses an 801-sample literature-grounded Choudhury valve-stiction benchmark with 3% deadband-plus-stickband and 1% slip jump. To avoid artificial statistical replication, one 201-point, 50-s cycle is used for quantitative identification, while the remaining deterministic cycles are retained only for repeatability visualization. Across ten repeated 70/30 train–test partitions, the adaptive neuro-fuzzy model achieved the lowest mean held-out NRMSE (0.001652), followed by the 16-rule (0.001894), nine-rule (0.002490), and fuzzy C-means (0.008684) models. In the full-cycle descriptive fit, the adaptive neuro-fuzzy model achieved the lowest NRMSE (0.001431), whereas the 16-rule model yielded the smallest loop-area error (0.0106%). Under 5% training-noise contamination, the compact fuzzy C-means model produced the lowest clean-test NRMSE, revealing an accuracy–robustness trade-off. The results demonstrate that path-parameterized fuzzy inference can provide an interpretable surrogate for valve-stiction-loop reconstruction, soft sensing, digital-twin visualization, and control-performance monitoring. Because the benchmark is model-generated rather than a raw plant acquisition, the results establish reconstruction fidelity and within-trajectory generalization rather than independent experimental prediction.

Abstract: We test whether round-number price levels predict reversals in gold, using 5,138 daily XAUUSD bars from 2005 to 2024. For every touch of a multiple of US\(100 we record whether the close returns to the side of the level on which the day opened, and we compare the round grid with 80 placebo grids of identical spacing shifted away from round numbers. Inference is based on the round grid's rank among all 81 grids and on a month-block bootstrap. Among 232 eligible first touches (231 usable same-day observations), the same-day holding rate is 45.0% (104/231) against 49.2% for the placebo grids; the round grid ranks 72nd of 81 (rank-based p = 0.247), and no primary test survives a Holm correction. Results are similar for US\)50 levels and for other first-touch windows. We find no robust evidence that round-number levels predict reversals under this daily-bar definition.

Abstract: Background. Toxic metals and essential minerals share transport, binding, and elimination pathways, so toxic-metal handling depends partly on mineral status. Interventions supplying a broad spectrum of minerals may therefore alter urinary toxic-metal excretion. Methods. Prospective, open-label, controlled pilot study. Experimental participants — a clinic population under treatment for chronic post-viral and post-vaccination illness — treated all drinking water with a biotite-derived sulfated ionic mineral solution at 5 mL per gallon and, as tolerated, added a concentrated oral dose of 1 mL per 20 pounds of body weight twice daily; healthy comparators received no intervention. Of 42 enrolled, 30 contributed paired specimens (20 experimental, 10 comparator) after withdrawals and the exclusion of six participants above the protocol age limit of 70 years. Urine panels covering 38 elements were obtained at baseline and follow-up (median interval 54 days, IQR 47–88). The protocol’s primary outcome was change in urinary toxic element levels; we compared changes from baseline between arms using the Mann–Whitney U test without multiplicity correction. A post hoc composite counted how many of six toxic metals — aluminum, arsenic, cadmium, lead, mercury, thallium — rose more than 15% from baseline; the threshold was fixed before that analysis and afterward a robustness check varied the threshold from 0 to 50% increases. Results. The protocol’s primary outcome — element-level change across the panel — was null: of the 38 elements measured, 29 had sufficient paired data for testing, and two reached nominal significance, aluminum (p = 0.024) and selenium (p = 0.018), against approximately 1.45 expected by chance; neither survives correction for multiplicity. The post hoc composite was not null. Experimental participants increased excretion across more toxic metals than comparators: a median of 3.0 of six versus 1.5 (p = 0.023, and p = 0.007–0.032 at every threshold from 0 to 50%). At least two metals rose in 95.0% versus 50.0% (p = 0.009). Applying the same threshold to six essential minerals, the arms did not differ (median 2.0 versus 2.0, p = 0.82). At post-study interview, eight of the 29 participants allocated to the experimental arm (27.6%) reported adverse effects: three were moderate to severe and required dose reduction or interruption; all resolved with adjustment, and none was serious. Five (17.2%) reported subjective improvement, and two more asked to continue treatment without describing a specific change; comparators were not interviewed. Conclusions. A biotite-derived sulfated ionic mineral complex was associated with increases in urinary excretion of more toxic metals than in contemporaneous comparators, while the number of essential minerals with increased excretion did not differ between arms. The direct within-person contrast favors selectivity for toxic elements by 15.8 percentage points (95% CI −10.3 to +41.5), an estimate this pilot is too small to resolve. The cause and mechanism of this change are uncertain, and urinary excretion alone cannot establish that body burden fell.

Abstract: We study the smallest nonnegative diagonal subtraction, supported on at most two sites, that makes the shifted discrete Dirichlet Laplacian positive semidefinite and singular. The size of the subtraction is its Euclidean amplitude norm, equivalently its Frobenius matrix norm. A comparison of Green matrices shows that the optimal sites are the central adjacent pair for even path length and either central adjacent pair for odd path length. For a reflection-symmetric pair, write its Green matrix as \(a\bigl(\begin{smallmatrix}1&q\\q&1\end{smallmatrix}\bigr)\). We give all global minimisers: equal amplitudes are optimal for q greater than or equal to 1/2, whereas two unequal reflected minimisers occur for 0 less than q less than 1/2. For a path of length 2m, this gives a unique shift threshold determined by \(\sinh((m+1)\theta)=2\sinh(m\theta)\); the thresholds increase to 1/2. We also give a quartic characterisation for unequal diagonal Green entries, prove strict improvement when an additional perturbation site is allowed, and derive fixed-shift and diffusive scaling limits. For any fixed support, we parametrise the first-contact boundary by a simplex of nonnegative sources. A thin QR factorisation represents the corresponding unit null vectors on a sphere, preserving their angles; three sites give an exact three-dimensional spherical triangle. We also derive the vector-ratio form of the two-site transition and a standard spectral-gap bound on null-vector sensitivity. Numerical calculations illustrate these statements without asserting a global three-site classification.

Abstract: DNA methylation biomarker development spans data discovery, assay-specific processing, marker selection, model development, and independent validation. Agentic systems could help coordinate these stages, but current studies use the term agent for systems with markedly different levels of autonomy and domain-specific evidence. We conducted a structured evidence synthesis, screening 2,801 unique records and organizing relevant systems along three dimensions: workflow autonomy (M0–M4), direct methylation-task evidence, and evidence maturity (E0–E5).Current systems cover complementary parts of the workflow. MethylCurate supports public methylation-data curation; AutoBA and BioMaster execute methylation-analysis pipelines; CpG Atlas provides a self-correcting query interface; and EpiBench, BixBench, and BixBench3 provide direct benchmark evidence for bounded methylation tasks. Across these studies, evidence is strongest for data access, workflow execution, and task-level evaluation. Among systems with task-level evidence reviewed here, we did not identify one that spans dataset discovery, marker selection, classifier development, independent validation, and sample-level uncertainty reporting within a single validated workflow.We therefore propose a matched evaluation framework comparing fixed workflows, rule-based routers, and general or domain-specific agents under shared data, tools, compute, and retry budgets. The framework separates execution reliability, scientific validity, replication, uncertainty, and resource use. Current evidence supports agentic components for methylation analysis, while end-to-end autonomous biomarker discovery remains an open evaluation problem.

Abstract: Pulsating heat pipes (PHPs) beneath a battery cold plate operate with the evaporator above the condenser, so that vapour expands directly against gravity. This orientation is imposed by vehicle packaging, yet its internal flow has rarely been resolved. A closed-loop borosilicate glass PHP of 2 mm/3 mm bore charged with methanol was visualized at 960 frames per second at 17, 22 and 27 W and charged volume fractions of 8.5, 17.0 and 25.6 %, and modelled with a one-dimensional force balance. Eleven flow regimes are grouped into three families-evaporation froth (A), oscillation confined to the condenser and adiabatic sections (B) and turn-to-turn migration (C)-and mapped in the power/filling-ratio plane. Vapour pressure dominates the force balance at 0.12 N, twice the wall friction, while gravity contributes less than 5 %. The model reproduces the measured plug amplitude to within 0.013 m at low filling ratio. Thermal resistance rises from 1.13-1.43 K/W at 8.5 % to 2.11-3.85 K/W at 25.6 %, so underfilling is favourable here. The map allows dry-out and flooding to be diagnosed from the observed regime in gravity-suppressed PHP battery modules.

Abstract: Introduction: Rotavirus remains the leading cause of severe gastroenteritis in young children worldwide despite widespread evidence of the positive impact of rotavirus vaccine. We evaluated the long-term impact of rotavirus vaccine introduction on rotavirus diarrhea hospitalizations in Togo. Method: We implemented active surveillance for all cause gastroenteritis in Togolese children < 5 years of age hospitalized at Sylvanus Olympio Teaching Hospital in Lome. Rotavirus testing with ELISA was performed at the sentinel site laboratory and genotyping performed at regional laboratories on stool samples collected from these children. To assess long term impact of rotavirus vaccine, we compared findings pre vaccine (2010-2014) and post vaccine (2015-2024) introduction. Results: Overall, 1,507 children hospitalized with all cause gastroenteritis were enrolled in the surveillance study, and 1,444 (95.8%) had rotavirus ELISA result, including 65.6% before rotavirus vaccine introduction. Of these 1,444 cases, annual median number of 189 (range=153-260) all-cause gastroenteritis hospitalizations before vaccine declined to 40 (range=27–71) after vaccine introduction, a 78.8% annual median reduction. Rotavirus-associated annual median prevalence decreased from 54.0% to 16.9% (range=5.3-41.3%) after vaccine introduction, and proportion rotavirus positive in < 1 year old infants decreased from 64.0% to 34.5%. The number of non-rotavirus gastroenteritis cases before and after vaccine introduction remained stable, while rotavirus cases became too small to allow appreciation of its seasonality after vaccine introduction. We observed a predominance of G1P[8] in 47.5% of cases, followed by G2P[4] in 10.2% of cases after vaccine introduction. Conclusions: We observed evidence of sustained impact of rotavirus vaccine, with a substantial decline in rotavirus gastroenteritis cases in Togo. Continuing surveillance will remain important to better monitor additional changes in rotavirus disease in Togo.

Abstract: Environmental and psychosocial conditions during pregnancy may influence neurodevelopment through interacting maternal and placental processes. This critical narrative review examines placental oxidative stress and mitochondrial regulation within a prenatal econeurobiological framework. Author-curated literature was supplemented by targeted bibliographic and original-source checks spanning environmental exposures, placental mechanisms, prenatal imaging, and childhood development. The synthesis distinguishes physiological redox signaling from oxidative injury and connects antioxidant defenses, trophoblast metabolism, and mitochondrial quality control with placental exchange and signaling. Experimental research supports selected links between maternal stress, placental oxidative responses, and offspring outcomes. Human biomarker and imaging studies provide complementary associations, including inconsistent and null findings, but do not establish placental oxidative mediation in humans. The framework distinguishes direct fetal exposure, altered placental function, and pathways involving obstetric complications, while retaining postnatal conditions as contributors to later performance. Its contribution is to specify which exposure–response relationships are supported and which require prospective testing. Temporally aligned biological sampling, developmental imaging, and functional follow-up are needed to determine when placental redox processes contribute to neurodevelopmental vulnerability. Antioxidant effects in experimental models motivate mechanistic investigation but do not establish clinical prevention of neurodevelopmental disorders.

Abstract: Missing values in materials datasets can compromise both data reconstruction and the prediction of material properties. However, an imputation method that accurately reconstructs individual descriptors does not necessarily maximise downstream predictive performance. Here, we investigate this distinction using 2,363 phosphor records from the Inorganic Phosphor Optical Properties database and compare three descriptor representations: a combined set of 88 features, a 13-feature physics-informed set, and a 75-feature data-driven set. Mean, K-nearest-neighbour, iterative Bayesian-regression and random-forest-based imputation methods are evaluated under artificially generated MCAR, MAR and MNAR mechanisms at missingness rates of 10–50%. Reconstruction error is assessed separately from emission-wavelength prediction using five regression algorithms and host-grouped cross-validation. In the evaluated configurations, random-forest regression performs best after feature-count selection, while the physics-informed representation approaches the predictive performance of the combined feature set. The iterative random-forest imputer achieves the lowest average normalised reconstruction error across all three representations. Nevertheless, the imputer minimising reconstruction error does not consistently maximise downstream predictive performance, with agreement between the two criteria in only 50.2% of the examined model-specific comparisons. These findings highlight the need to evaluate imputation according to its intended use rather than relying on reconstruction metrics alone. The reported performance remains conditional on the descriptor-selection procedure, host grouping and simulated missingness mechanisms; independent validation is required before extrapolation to chemically novel phosphors.