Abstract: Growing consumption of packaged products through retail and e-commerce is increasing the amount and complexity of packaging waste entering urban waste systems. The challenge is especially important in rapidly growing cities where limited source separation, collection constraints, fragmented recovery activities, and dependence on informal recycling can restrict material circularity. This study develops a qualitative system dynamics framework for circular urban packaging waste management, using Dhaka, Bangladesh, as the principal developing-city context and Chicago, USA, as a limited structural reference. Published literature and secondary evidence are synthesized to characterize relationships among waste generation, consumer separation, collection, reverse logistics, recycling capacity, and secondary-material recovery. A causal loop framework identifies four reinforcing and two balancing feedback loops, and four structured what-if conditions—current trajectory, worst case, moderate transition, and best case—are examined qualitatively. The analysis indicates that isolated interventions may shift system bottlenecks rather than resolve them. A coordinated transition requires packaging prevention, reliable source separation, integration of formal and informal reverse flows, aligned sorting and recycling capacity, and viable demand for recovered materials. The framework provides a theoretical basis for coordinated urban circular-economy policy and future quantitative system dynamics modeling.

Abstract: Background: Stroke prediction using machine learning is crucial for early detection, enabling timely interventions to reduce mortality and long-term disability. Our study aimed to find the best scaling method and classifier for stroke prediction. Methods: We applied three scaling techniques to the datasets: standardization, minmax, and robust scaling. For each scaling method, we employed eight classifiers: Gaussian Naïve Bayes, Decision Tree, Support Vector Machine, Logistic Regression, Random Forest, K-Nearest Neighbors, Extreme Gradient Boosting (XGBoost), and Bagging Method. We used performance metrics such as accuracy, F1 score, recall, and precision to evaluate the model's performance. 10-fold cross-validation was conducted, and accuracy, precision, recall, and F1 score were assessed to determine the most robust classifier for the dataset. All analyses were performed using Python. Results: Across all the scaling techniques, XGBoost delivered remarkable performance. For standardization, min-max, and robust scaling, XGBoost achieved accuracies of 95.44%, 95.56%, and 95.39%, respectively. The Bagging and Random Forest Classifier also demonstrated strong accuracy across various scaling methods. Despite no statistical difference among the scaling methods, min-max scaling demonstrated a slight performance improvement. Under min-max scaling, the Bagging Classifier reached an accuracy of 93.32%, whereas the Random Forest Classifier attained an accuracy of 94.53%. Decision Tree 90.73%, SVM 83.70%, Logistic Regression 83.35%, and K-NN 88.02%. Conclusion: XGBoost achieved the highest accuracy, while min-max scaling enhanced the performance of all classifiers. These results highlight the importance of scaling and the effectiveness of XGBoost for stroke prediction.

Abstract: This paper presents PySkyLumos, an open-source Python framework for simulating skylight polarization measurements and metadata. It implements the Rayleigh, Berry, and Pan models, capturing both classical scattering and elevation-expressed singularity dynamics. Efficiency benchmarks show a fivefold overall speedup and up to two orders of magnitude faster Stokes-parameter image formation compared to existing tools. Validation against real polarization camera data confirms its accuracy and suitability for bio-inspired computer vision, navigation, robotics, and sensing applications.

Abstract: Hormone-driven cancers, including prostate, breast, endometrial and ovarian cancers are a leading cause of cancer-related mortality. Endocrine therapies targeting androgen receptor and oestrogen receptor/progesterone receptor signalling have improved outcomes, but acquired resistance arising through genetic and epigenetic mechanisms that bypass receptor dependence remains a major challenge. BORIS (Brother of the Regulator of Imprinted Sites), a CTCF paralogue, is a germline-restricted regulator that becomes aberrantly reactivated in cancer through promoter hypomethylation, p53 loss, and CTCF displacement. Once expressed, BORIS displaces CTCF at shared genomic binding sites, disrupts TAD boundaries and reprograms CTCF binding sites into alternative transcriptional start sites, through SRCAP-mediated H2A.Z deposition, to activate germline genes that sustain proliferation, and therapeutic resistance. BORIS expression correlates with poor prognosis in ovarian and endometrial cancers, associates with ER/PR levels in breast cancer, and correlates with Gleason score in prostate cancer. BORIS represses AR signalling in ovarian cancer and inversely correlates with receptor expression in endometrial cancer. Its cancer-testis antigen properties position BORIS as an immunotherapy target, with preclinical vaccine efficacy demonstrated in breast cancer models. This review synthesises disease-specific evidence positioning BORIS as a driver of transcriptional reprogramming and endocrine resistance, identifies current knowledge gaps, and discusses translational implications of targeting this pathway.

Abstract: Krabbe disease (KD) is an autosomal recessive neurodegenerative lysosomal disease typically caused by pathogenic variants in GALC, which encodes for the lysosomal galactosylceramidase enzyme. Age of onset is variable, with earlier symptom onset conferring a more severe, and ultimately fatal, disease course. We present a case of an 18-month-old previously healthy female diagnosed with late-infantile KD following a 3-month period of psychomotor regression and irritability. Clinical features included severe axial hypotonia with lower extremity spasticity and hyperreflexia. Whole genome sequencing disclosed a complex GALC genotype including two likely pathogenic variants, p.(Gly284Ser) and p.(Tyr314del), which had not been observed in compound heterozygosity prior. Magnetic resonance imaging displayed symmetric T2 hyperintensity of the central-parietal white matter. The patient ultimately died 6 months after diagnosis from rapidly progressing neurological deterioration. There was significant atrophy on neuropathologic examination, with diffuse globoid cell infiltrate at the supratentorial and infratentorial white matter with relative sparing of the frontal lobe on light microscopy. This patient’s novel genotype, imaging and pathologic findings expand our understanding of the phenotype-genotype correlation in individuals with KD.

Abstract: Brassica napus L. is a member of the family Cruciferae. This species is also referred to as "rape," “oilseed rape," "rapa," "rappi," and "rapeseed." It is the third most important source of edible oil globally after soybean and palm oil. Quantitative trait loci (QTLs) analysis has proven to be a powerful tool to reveal the genetic mechanisms of seed weight regulation in this species. We identified and compiled a comprehensive list of 1,905 candidate genes associated with seed weight regulation in Brassica napus. Five candidate genomic regions with 204 seed weight–related genes were identified, including 21 differentially expressed genes from transcriptomic analysis. A CRISPR/Cas mutant library targeting silique- and seed size-related genes was generated in rapeseed, indicating a potential role of BnaHRDs in their development. DEGs were involved in developmental processes, cell division, and nutrient storage, indicating their role in regulating seed size in Brassica napus. HB1 and HAIKU2 also regulate seed weight by controlling endosperm proliferation through the IKU pathway. A total of 1,643 SNP markers were aligned to the pseudochromosomes of Brassica napus using BLAST to identify significant hits. Furthermore, BRASSINAZOLE-RESISTANT1 (BZR1) is an important regulator of seed growth acting through maternal tissues. TDZ treatment enhanced both cell size and number. Plant hormones, including auxin, gibberellin (GA), and brassinosteroid (BR) signaling pathways, have been reported to regulate seed size. Transcription factors including LEC2, WRI1, FUS3, MYB30, and ABI3 were active during early seed development, while LEC1, LEC2, ABI3, and FUS3 remained active during seed maturation. Physiological processes play a key role in yield variation and the selection of high-yielding genotypes. Currently, marker-assisted selection and transgenic approaches are commonly used in rapeseed breeding. Temperature and light are key environmental factors affecting oil production in rapeseed. The integration of genetic engineering, molecular marker technologies, and conventional selection is expected to be a key approach for improving high-oil-content rapeseed. These limitations highlight the need for further studies on rapeseed applications.

Abstract: This paper undertakes a philosophical investigation of the relationship between meta-laws, first-order laws, and observational conditions, and provides a categorical formalization of that relationship. We begin by establishing a four-quadrant ontology that makes explicit the possible configurations of relations among M, L, and O. We then propose a theory of ontological degeneration, arguing that the remaining three quadrants may be regarded as specializations of quadrant IV-C, obtained by contracting O or L to a terminal object; under an interpretive framework that preserves the explanatory power of scientific practice, these specializations may reasonably be called “degenerate forms.” We further prove that symmetry, normativity, causality, and statistical covariance can be derived only from quadrant IV-C, and from none of the other quadrants. Against this ontological background, the central thesis of the paper is that F(L,O) = Ψ(R(L,O)) is the local projection or reconstruction of the global meta-law M at the specific pair (L,O). From this thesis we derive seven successive conclusions: M ∼ F(L,O); Sym(M) = Sym(F); normativity is the constraint structure of covariance; causality is the functoriality of F over the category of time; statistical covariance is the functoriality of F over the category of probability; historicity is the functoriality of F along the flow of time; and modality is the functoriality of F over the group of admissible transformations. We further raise F to a 2-functor, establish the higher-categorical structure of F, and embed F into topos theory, constructing the F-topos, in which the conditionality, boundedness, hierarchy, symmetry, normativity, causality, modality, and higher structure of laws can all be precisely characterized in the internal logic of the topos. We also argue that the F-framework, though a meta-theory and thus not directly falsifiable, generates three classes of empirically testable assertions: covariance predictions, obstruction predictions, and levelevolution predictions. Finally, we draw out the methodological implications of the framework and engage it in dialogue with phenomenology, pragmatism, and critical theory.

Abstract: Detecting tiny defects on printed circuit boards (PCBs) requires a balance among preserving high-resolution details, fusing semantics across scales, and maintaining deployment efficiency. This paper presents ASBNet-DFD, a lightweight defect detection network designed for this setting. Instead of repeatedly upsampling, downsampling, and concatenating features at their native resolutions, ASBNet-DFD concentrates shared semantic interaction from different feature levels in a fixed low-resolution space. The resulting shared context is then injected into three detection scales through scale-specific gates, reducing repeated cross-scale computation on high-resolution feature maps. Because the highest-resolution detection branch requires both local detail and semantic context, the network further introduces an asymmetric semantic bridge that provides a single directed semantic compensation from the adjacent medium-resolution backbone feature to this branch. This adds context for fine-grained detection while keeping the extra computation limited. With 640×640 input, ASBNet-DFD has 0.873 M parameters and requires 5.1 GFLOPs. Across seven independent training runs on DeepPCB, the validation mAP50–95 is 0.8109±0.0016, indicating low run-to-run variation under the fixed training protocol. Under a unified RTX 4070 and TensorRT FP16 deployment setup, the model achieves an mAP50–95 of 0.6919 on the independent test set with a network forward latency of 1.175 ms/image, lower than YOLO11n, YOLO12n, and YOLO26n under the same test protocol. Under the current quantization configuration, the INT8 mixed-precision implementation provides neither a speed benefit nor a storage benefit. These results show that, under the experimental conditions used here, concentrating most cross-scale semantic interaction in a low-resolution domain while retaining only the necessary directed semantic transfer to the highest-resolution branch can provide competitive detection performance and practical inference efficiency with a compact model.

Abstract: This work presents a soft finger designed, simulated, and experimentally tested, combining pneumatic actuation, variable stiffness, and embedded sensing. The finger integrates a fiber-reinforced pneumatic actuator that generates single-input underactuated bending through three interconnected chambers located at the MCP, PIP, and DIP finger joints. The design methodology follows bioinspired principles and cost-effective manufacturing constraints, producing a prototype via silicone casting with 3D-printed molds and incorporating a bio-based granular substrate for variable stiffness that allows post-deformation stiffening without compromising flexibility. Bending under pressurization is simulated utilizing the Finite Element Method (FEM) in ANSYS software, with the finger modeled using hyperelastic and fiber-composite material properties. Comparison of simulated and experimental results reveals an average deviation of approximately 8°. The adjustment of stiffness properties is achieved via vacuum-induced granular jamming, increasing the applied forces by up to 60% relative to the unstiffened state. An embedded sensing system (microphone) in the fingertip enables discrimination of contact surfaces based on captured vibration signals. These results indicate that finger deformation and stiffness can be controlled with minimal hardware and a simple design, supporting the use of this finger as a component in soft grippers or robotic hands.

Abstract: Wind resource assessment increasingly depends on integrating heterogeneous reanalysis, turbine data, geospatial operations, and persistent computational services. This article presents WindAnalysis, a containerized web framework for coastal wind studies in the Colombian Caribbean. Its novelty is architectural and operational: rather than introducing new wind models, the framework integrates scheduled ERA5 and MERRA-2 ingestion, persistent PostgreSQL storage, typed FastAPI services, an authenticated React geospatial interface, experiment retention, turbine-level power curves, and inverse-distance-weighted (IDW) interpolation within a single reproducible workflow. The evaluated database state comprised 32,633,302 ERA5 and 3,941,280 MERRA-2 records. Controlled benchmarks showed a 3962-fold acceleration for planner-based row counts and a 19.4-fold acceleration for 1% block-sampled mean queries, with 0.39% mean absolute relative error. Internal leave-one-grid-point-out IDW cross-validation yielded a 10 m wind-speed RMSE of 0.366 m/s for the deployed p = 2, 50 km configuration. The documented La Guajira case, using direct ERA5 wind at 100 m, produced a gross AEP of 8.901 GWh/year and a gross capacity factor of 61.58% for a Vestas V82 turbine. The evaluation further includes 2839 archived automated tests and release-level UI-to-runner parity. WindAnalysis therefore provides an inspectable, repeatable integration layer for regional wind-resource assessment while keeping computational reproducibility distinct from external physical validation.

Abstract: We study the Vertex Shift Method (VSM): given a polynomial P with a real critical point φ (i.e. P′(φ) = 0), the translation Q(y) = P(y + φ) eliminates the linear coefficient of Q exactly (Theorem 1) and induces a similarity transformation on the companion matrix (Proposition 1). The supporting theory is deliberately elementary – Taylor expansion, a standard non-derogacy argument, and a direct eigenvalue estimate – and is developed only as far as needed to support three specific, verified strengths of the resulting preconditioning procedure. First, VSM avoids a structural barrier that matrix balancing cannot: for singular companion representations (e.g. continuous-time Markov generators, where the constant coefficient c₀ = 0 by conservation of probability), balancing provably cannot help, since diagonal similarity preserves c₀ exactly, while VSM’s translation changes it, moving all 200/200 tested instances from singular or extremely ill-conditioned to finite, moderate condition numbers. Second, VSM composes multiplicatively with tropical scaling (Noferini et al.): the two act on different orbits of the representation – one an exact local translation, the other a global geometric rebalancing – and their composite achieves gains of 2.9 × 10¹¹× on Wilkinson W₂₀, orders of magnitude beyond either alone, a result we support with two analytical bounds and a corrected, densely-verified numerical check across degrees n = 3 to 20. Third, applied to the Kerr black hole equatorial radial potential, with coefficients reconstructed exactly from cited primary-source physical parameters, VSM gives a verified 14.48× conditioning improvement (25.02× combined with balancing) on a fully specified 100-point spin sweep. Beyond these three results we give exact-sweep, proxy, and hybrid critical-point selection rules with characterized costs, an implicit-function-theorem extension to time-varying polynomials with a documented and partially resolved branch-switching failure mode, and a systematic mapping of the minimum eigensolve precision (100–112 bits) needed to prevent conditioning gains from being offset by floating-point cancellation in root recovery – a divergence we document explicitly on Runge- and Kahan-type instances rather than let κ(C_Q) stand as a proxy for accuracy. We report two claims from earlier stages of this work that do not reproduce under independent re-implementation (a Monte Carlo forward-error win rate and one curved-shift adaptive-policy tracking figure) as open, unresolved discrepancies (Section 8.1) rather than either defend them uncritically or silently remove them. A third figure, the curved-shift hybrid-10/20 tracking result, was also flagged as discrepant in an earlier draft; that was traced to a bug in independent verification tooling rather than a reproducibility gap in this paper’s own claim, and is no longer listed here (Section 8.1). The central contribution of this paper is to establish critical-point translation as a mathematically analyzable preprocessing operation for companion-matrix polynomial eigenvalue problems. An earlier version of this claim described VSM’s relationship to balancing as general “complementary behavior”; that phrasing is retired here, for the same reason a companion paper in this research line retired it independently, after an audit found it overstated in a high-degree regime map. VSM has a structural advantage balancing provably cannot share in one specific, proven regime – singular companion representations with c₀ = 0, where diagonal similarity preserves c₀ exactly and VSM’s translation does not (Section 7.2) – and the two methods act on different orbits of the representation (translation vs. diagonal similarity) elsewhere. This paper does not claim general equal-footing complementarity between VSM and balancing across all regimes; where they were compared head-to-head (Section 7.1’s Kerr sweep, corrected), VSM alone was found competitive with, not dominated by, balancing alone in 66.0% of trials on that application.

Abstract: The Year 2026 was declared the International Year of Rangelands and Pastoralists (IYRP 2026) by the United Nations General Assembly; it emphasizes the global concurrence to protect these important ecosystems. More than 50% of the Earth's terrestrial area is covered by rangelands that host millions of pastoralists despite serious challenges due to overgrazing, climate change and variability, coupled with changes in land use patterns. Addressing these challenges requires integrated science-based management and effective policies to enhance the long-term ecological sustainability and productivity of rangelands. This review paper offers an extensive review of rangeland management, particularly focusing on the Kingdom of Saudi Arabia (KSA), and places Saudi rangelands in a global context, highlighting the primary factors contributing to rangeland deterioration, such as socioeconomic transformation and traditional nomadic practices disappearance. The study investigates best practices of rangelands management in the global. Saudi Arabia's rangelands cover around 146 million hectares, represent about 73% of the country's total area, and sustain more than 31 million livestock. Unfortunately, almost 70% of these fragile ecosystems are experiencing moderate to severe degradation. The review paper concluded by proposing a set of policy interventions designed to promote long-term environmental sustainability, in accordance with achieving the Saudi Green Initiative and the global Sustainable Development Goals (SDGs) targets such as national afforestation program, national rangelands strategy and governance and regulating camels raising farms and livestock grazing systems.

Abstract: This paper reports a document-based, multiple-case study of six Internet-of-Things (IoT) prototypes designed and simulated during a one-week immersive course, “IoT for Data Intelligence,” delivered in July 2026 within the professional Master in Applied Artificial Intelligence (Maestría en Inteligencia Artificial Aplicada, MNA) at Tecnológico de Monterrey. Six teams followed the same five-day toolchain IoT theory; Oracle Application Express (APEX), SQL, and REST service design; MIT App Inventor; ESP32/Wokwi simulation; and generative-AI integration and produced Wokwi-simulated prototypes spanning industrial energy monitoring, agricultural hazard response, residential automation, cardiovascular telemonitoring, industrial waste reduction, and precision agriculture. A fixed coding framework was applied across architecture, AI-integration pattern, platform-level failure modes, security debt, and Sustainable Development Goal alignment, distinguishing findings that the course structure itself prescribes from findings the teams introduced independently. The six cases converged on a shared five-layer architecture and, in a pattern only partly prescribed by the course, on keeping generative AI in an advisory or fail-safe-wrapped role. Deposited results were also compared, for illustrative purposes only, against the course’s internal competency rubric. An observed proposal from a Pontifical Catholic University of Chile’s collaboration is discussed as an informal reference point rather than as evidence for generalization. This paper discusses the implications and limits of this small, single-institution, single-cohort, simulation-only case set.

Abstract: Survival analysis has been widely used to understand the etiology of chronic diseases. However, traditional survival analysis models only a single endpoint, whereas the development of chronic diseases is a multi-state process. In this review, we examine the connections among logistic regression, the Cox model, competing risk models, and multi-state models in estimating hazards and survival risks, with findings summarized in three aspects. First, logistic regression and the Cox model are connected. The conditional likelihood of a conditional logistic regression stratified by risk set is equivalent to the partial likelihood used by the Cox model. The survival risks can be derived from the Cox model using the Breslow estimator. Alternatively, pooled logistic regression can be used to estimate risk within small time interval that approximate discrete-time hazard, with survival risks estimated using the Kaplan-Meier (K-M) estimator. Survival risk estimated from the Cox model is more accurate, whereas discrete-time hazard is more flexible for creating complex statistics (such as counterfactual survival risks in causal inference) and provides an approach for integrating machine learning into survival analysis. Second, for nonparametric estimation of survival risks from hazards, the cumulative incidence functions (CIFs) used in competing risks and the Aalen-Johansen (A-J) estimator used in multi-state process are extensions of the K-M estimator used in single disease endpoint. Third, the cause-specific Cox model and the Markov Cox model are extensions of the Cox model to competing risk and multi-state settings, respectively. Correspondingly, multinomial pooled logistic regression and a discrete-time split-state framework extend pooled logistic regression to competing risk and multi-state settings. Our paper can serve as a tutorial to illustrate the connections between survival analysis and multi-state modeling. We anticipate that multi-state models will play an increasingly important role in understanding chronic disease dynamics and advancing precision prevention and prediction.

Abstract: Background/Objectives: The VA Emergency Medicine Addiction Hotline (VEMAH) is a real-time telehealth addiction consult service for Veterans Health Administration (VHA) providers in acute-care settings. VEMAH launched in May 2023 across eight Veterans Integrated Service Network 22 (VISN 22) facilities, expanding to 17 total VHA sites over 3 years. Results: Over 38 months (May 7, 2023 through July 23, 2026), the service completed 2,154 encounters, including real-time e-consults, direct patient tele-care, care coordination, proactive tele-outreach, and provider education. The most common indications for consultation were opioid- and alcohol-related concerns. Pain-related concerns accounted for about 11% of all encounters. VEMAH providers focused on proactive outreach, substance use disorder (SUD) pharmacotherapy interventions, and coordinating care with local SUD services for ongoing treatment. Conclusions: We describe the feasibility of implementing and scaling a real-time telehealth addiction consult service across VHA acute care settings. A descriptive comparison of ED OSI dashboard trends before and after VEMAH implementation showed positive concurrent trends in naloxone and buprenorphine prescribing within VISN 22, though these observations are hypothesis-generating rather than confirmatory. These findings support VEMAH as a scalable model for extending addiction care access across VHA acute-care settings.

Abstract: The rapid ascent of Large Language Models (LLMs) has empirically demonstrated that complex syntactic reasoning and semantic manipulation do not require subjective experience (qualia). This paper proposes the Ontological Interface Hypothesis (OIH), a triadic framework that decouples cognition, consciousness, and biological self-preservation. We argue that while cognition is an algorithmic, substrate-independent process susceptible to the Data Processing Inequality and entropic degradation (Model Collapse), consciousness and biological negentropy may involve non-algorithmic aspects of a more fundamental ontological layer. We model the thalamocortical complex not as an additional computational engine, but as a phase-synchronized biological interface connecting the algorithmic substrate to this deeper realm. Furthermore, we address standard functionalist counter-arguments, suggesting that artificial agents lacking such an interface are likely to remain prone to cognitive drift and physical degradation, functioning primarily as deterministic proxies of human intent rather than autonomous subjective entities.

Abstract: The aim of the present study is to analyse the relationship between sociodemographic variables and the perception of obstetric violence in Spain. A cross-sectional study was carried out through an online questionnaire. A descriptive analysis of all variables was performed followed by a bivariate analysis using the Chi-Squared test. Finally, all statisti-cally significant variables were included in a multivariate binary logistic regression mod-el. Data were processed and graphs created using Jamovi 2.3.28 software. Statistical sig-nificance was set at p< 0.05. A total of 6060 women were included in the study. Some 33.5% (n=1578) referred to having experienced obstetric violence or not being aware of it. Variables that showed a statistically significant relationship with obstetric violence were (p< 0.05): aged between 18 and 25 years, being a student, being of Romani ethnicity, being treated in public health, and labour being completed with urgent caesarean section or in-strumental delivery. The binary logistic regression model proved significant (X2=314; p< 0.001), with high sensitivity (0.970), an AUC=0.647 and a Nagelkerke pseudo-R2 deter-mination of 0.079. Young women face higher obstetric violence, influenced by occupation and ethnicity, reflecting socioeconomic and cultural care impacts. Improving obstetric practices is crucial for women's autonomy and respectful childbirth care.

Abstract: The metabolic landscape of the post-implantation mouse embryo and its associated placenta remains incompletely characterised. Here, we performed untargeted LC–MS-based metabolomics on mouse embryos from E9.5 to E13.5 and on placentas from E10.5 to E12.5, and integrated these data with a published single-cell transcriptomic atlas of the same developmental window. We found that the embryonic metabolome undergoes progressive remodelling, with the most pronounced changes occurring around E11.5. By combining time-course analysis, WGCNA-based network prioritisation, and embryo–placenta comparison, we identified 56 core metabolites that increased coordinately in both tissues during this period. Single-cell analysis revealed that the definitive erythroid lineage was enriched for nicotinate and nicotinamide metabolism, with progressive upregulation of the NAD+ biosynthetic enzyme Nmnat2 during erythroid maturation. Correspondingly, niacinamide, NAD, and 1-methylnicotinamide increased in the embryo from E10.5 to E13.5. Regulon analysis showed that haematopoiesis-related transcriptional programmes, including Spi1 and Gfi1, correlated with development-associated metabolic modules. Together, our data provide a stage-resolved metabolomic resource and identify candidate metabolites that may support embryo growth in vitro.

Abstract: Journal Entry Testing (JET) is a fundamental audit procedure to identify any potential misstatements, fraud, and management override of controls. Traditional rule-based JET methods suffer from high false positive rates and limited ability to detect complex anomaly patterns. Recent work has shown that large language models (LLMs) can serve as effective anomaly detectors for bookkeeping data, but LLM-only approaches lack explicit modeling of accounting constraints and structural relationships among entries. We propose the Constraint-Guided LLM-GNN (CG-LGN) framework, which integrates three complementary modules: (1) a heterogeneous graph neural network (GNN) that models structural relationships among journal entries, accounts, users, and temporal attributes; (2) an accounting constraint module encoding domain-specific rules such as debit-credit integrity, unusual account combinations, and period-end concentration; and (3) an LLM-based explanation generator that produces auditor-readable interpretations for flagged entries. Experiments on synthetic journal entry data with six injected anomaly types show that CG-LGN achieves a PR-AUC of 0.49 (a 0.20 absolute gain over the strongest single-module baseline) and reduces false positives per 1,000 entries by 47%. Ablation studies confirm that the GNN and constraint modules improve detection performance, while the LLM module improves explanation quality.

Abstract: Background/Objectives: Preschool teachers can reinforce oral health practices, but the persistence of educational messages requires evaluation. This study assessed changes in knowledge, beliefs and self-reported practices between 3 and 12 months after school-based education. Methods: Education reached 240 teachers in 63 schools in Trabzon, Turkey. Questionnaires were completed by 222 teachers at 3 months and 152 at 12 months. Paired analyses included the 152 teachers assessed at both follow-ups. Responses were compared using McNemar tests. No matched pre-education assessment or concurrent control group was available. Results: The paired cohort represented 63.3% of trained teachers. Endorsement of fluoride-containing products for caries prevention decreased from 63.8% to 40.1%, recognition of enamel strengthening from 82.9% to 66.4%, and identification of first-tooth eruption as the time to begin cleaning from 92.8% to 75.7% (all p < 0.001). Recognition that avulsed primary teeth should not be replanted decreased from 83.6% to 70.4% (p < 0.001), whereas endorsement of immediate permanent-tooth replantation changed little (91.4% versus 89.5%; p = 0.508). Twice-daily brushing decreased from 82.9% to 73.0% (p = 0.001). Tests were exploratory and unadjusted for multiple comparisons. Conclusions: Several preventive messages were less frequently endorsed at 12 months. Follow-up identifies priorities for updated content and refresher education concerning fluoride, early oral care and primary-tooth avulsion. These observations do not establish initial educational effectiveness or clinical benefit in children.