Abstract: Introduction: Stunting is associated with poor nutritional intake during early childhood. This study evaluated the effect of a daily intake of 510 mL of an oral nutritional supplement for 180 days on linear growth among children with stunting and at-risk of stunting aged 12–36 months. Methods: A community-based, single arm intervention was conducted among 91 children in Kelantan, Malaysia. The children at enrolment had height-for-age Z-scores (HAZ) between <-1.0 SD to >-3 SD based on WHO Growth Standards. Anthropometric measurements were taken at baseline (T0), day-90 (mid-intervention), and day-180 (post-intervention). Nutrient intake was assessed using 24-hour dietary recalls, and compliance was monitored via returned empty sachets. Results: The mean age of the children at baseline was 26.7 ± 6.5 months, with 37 (41%) being stunted and 54 (59%) were at-risk of stunting. After intervention, the linear growth (height-for-age Z-score) was significantly improved over time (p < 0.001) in both stunted and at-risk children. A significant time-by-group interaction (p = 0.014) indicated differential effects between the stunted and at-risk groups. Post-hoc analysis showed HAZ improvements from baseline (T0) to 180 days in stunted and at-risk groups (p < 0.001) with the stunted group having greater mean differences. The number of stunted children declined by 37.8% (p = 0.003). Nutrient intakes of protein, vitamins D, vitamin C, vitamin B-complex, calcium, phosphorus, magnesium, and iron improved significantly. Conclusion: Daily intake of 510 mL of oral nutrition supplement improved linear growth and nutrient intakes. These findings support the potential of targeted supplementation in addressing child growth faltering and micronutrient inadequacies.

Abstract: Conventional image stitching pipelines predominantly rely on homographic alignment, whose planar assumption often breaks down in dual-camera configurations capturing non-planar scenes, leading to visual artifacts such as geometric warping, spherical bulging, and structural deformation. To address these limitations, this paper presents SENA (SEamlessly NAtural), a geometry-driven image stitching approach with three complementary contributions. First, we propose a hierarchical affine-based warping strategy that combines global affine initialization, local affine refinement, and a smooth free-form deformation field regulated by seamguard adaptive smoothing. This multi-scale design preserves local shape, parallelism, and aspect ratios, thereby reducing the hallucinated distortions commonly associated with homography-based models. Second, SENA incorporates a geometry-driven adequate zone detection mechanism that identifies parallax-minimized regions directly from the disparity consistency of RANSAC-filtered feature correspondences, without relying on semantic segmentation or depth estimation. Third, building upon this adequate zone, we apply anchor-based seamline cutting and segmentation, enforcing one-to-one geometric correspondence between image pairs by construction and reducing ghosting, duplication, and smearing artifacts. Extensive experiments on challenging datasets demonstrate that SENA achieves alignment accuracy comparable to leading homography-based methods, while providing improved shape preservation, texture continuity, and overall visual realism.

Abstract: Background/Objectives: Positive Psychotherapy (PPT) is an empirically supported treatment that directly targets positive resources and personal strengths as its primary logic. PPT is effective in amplifying happiness and well-being as an additional way to enhance positive mental health while also in ameliorating symptoms of negative affect, especially in depression, anxiety disorders and stress disorders. However, only few studies have been conducted to investigate these effects in the long run. Methods: This study examined the long-term outcomes of PPT in comparison to an active control treatment, Cognitive Behavioral Therapy (CBT). Forty-nine out-patient participants with a DSM-IV diagnosis for depressive disorder (MDD, Dysthymia) were treated with 14 sessions of manualized PPT (n = 23) or CBT (n = 26) group therapy. In a randomized controlled 2-center-study, questionnaires on depressive symptoms (BDI-II, MADRS, DHS), psychological distress (BSI), and well-being related outcomes (FS, PPTI, SWLS) were administered at baseline and 18 months follow-up. Results: Analyses using linear mixed models indicated significant differences in long-term treatment outcome for depressive symptoms (BDI-II, DHS, MADRS) and satisfaction with life (SWLS) depicting better outcomes for the PPT group. Between group effect sizes at 18 months follow-up were primarily in the middle range for all outcome measures, in favor of PPT. Conclusions: This study provides support for the long-term efficacy of PPT in the treatment of depression and improvement of positive resources.

Abstract: Background: Cellulite is a highly prevalent condition with dermal and subcutaneous alterations poorly captured by visual grading systems. Ultrasound has emerged as a non-invasive imaging modality capable of objectively quantifying morphological fea-tures relevant to cellulite. This systematic review evaluated the evidence on ultrasound for the diagnosis, structural characterization, and treatment monitoring of cellulite, identifying methodological limitations and research gaps. Methods: This systematic re-view (PROSPERO:CRD420251185486) followed PRISMA statement. Searches were conducted in PubMed, Scopus, and CENTRAL up to November 2025. Risk of bias was evaluated using ROBINS-I and the Newcastle–Ottawa Scale. Results: Nine studies involving 785 participants were included. Ultrasound frequencies ranged from 12 to 35 MHz, with some scanners operating across broader bandwidths. Despite variability in devices, acquisition protocols, and clinical comparators, all studies consistently demonstrated that ultrasound quantifies key structural characteristics of cellulite. Diagnostic investigations reported moderate-to-strong correlations (r ≈ 0.31–0.64) between ultra-sound-derived measures and clinical severity scores. Interventional studies showed measurable reductions in dermal and subcutaneous thickness, decreased adipose protrusion height, and improved dermal echogenicity across multiple treatment modalities. Ultrasound frequently detected microstructural remodeling not readily visible on clinical examination. Conclusions: Ultrasound is a valuable imaging modality for objectively characterizing cellulite and monitoring treatment-induced tissue remodeling. Standardized acquisition protocols, validated analytic criteria, and larger controlled studies are needed to support integration into routine dermatologic and aesthetic practice. The quantitative and reproducible nature of ultrasound-derived parameters also provides a suitable foundation for future integration with data-driven and artificial intelligence–based image analysis frameworks.

Abstract: Milk is a primary source of vital nutrients and bioactive components fundamental for the growth and development of both newborn animals and humans. Produced by economi-cally significant livestock species (including cattle, buffaloes, goats, sheep and camels) milk is a complex matrix rich in caseins, vitamins, fats and proteins. In addition to its nu-tritional profile, milk serves as a vehicle for milk-derived extracellular vesicles (mEVs), a specialized class of food-derived EVs (fEVs) that exert pleiotropic effects aligned with the One Health concept, relating animal health, human nutrition, and ecosystem stability. mEVs offer unique advantages, such as high biocompatibility and gastrointestinal stabil-ity, rendering them also potential therapeutic tools, as drug delivery systems. However, challenges remain regarding the standardization of mEVs and the variability of their mo-lecular cargo. This review provides a comparative analysis of mEVs across diverse spe-cies, including bovines, water buffaloes, yaks, camels, goats, pigs, horses, donkeys, and humans, with a focus on their unique functional profiles. Indeed, a critical issue in mEVs research is the isolation process: recommendations to minimize contamination from milk fat globules and casein micelles (which can cover EV signals) are given. Finally, current detection methods and instrumentation, with a specific focus on advancing Flow Cytom-etry (FC) approaches are discussed. Key insights include the use of Conventional FC (with fluorescence triggering, the necessity of rigorous controls and calibration, and the utility of Bead-Based Assays to overcome resolution limits) and of Imaging Flow Cytometry (IFC). In both technical approaches, the application of different EVs generic fluorescent markers and the strategic selection of tetraspanins (i.e. CD9, CD63, CD81), is mandatory, empha-sizing the importance of selecting appropriate antibody clones or considering cross-reactivity when targeting these antigens across different mammalian species.

Abstract: Sustainability has become a strategic priority for Higher Education Institutions (HEIs), particularly in the context of the Sustainable Development Goals, where university research plays a key role in addressing environmental, social, economic, and institutional challenges. However, the evaluation of sustainability-oriented research models remains limited by fragmented indicators, descriptive approaches, and the absence of robust, data-driven assessment frameworks. This study proposes a comprehensive framework for assessing the sustainability orientation of university research models, integrating validated measurement instruments with advanced analytical and predictive techniques to support evidence-based decision-making in higher education governance. The framework is based on a multidimensional instrument comprising 26 indicators across environmental, social, economic, and institutional dimensions, developed through expert judgment using the Delphi method and statistically validated by Confirmatory Factor Analysis (CFA). The instrument was applied to 260 researchers from four public HEIs located in the Colombia–Ecuador border region, and perceived performance was contrasted with actual institutional indicators, revealing significant non-linear discrepancies. To address this complexity, an artificial neural network model was developed to estimate real sustainability performance based on survey data, achieving a predictive accuracy of 90.92%. Beyond institutional diagnosis, the proposed framework functions as a decision-support tool that enables HEIs to identify critical gaps, prioritize interventions, and guide continuous improvement strategies in research management. Due to its methodological rigor, scalability, and transferability, the framework can be adapted to different higher education contexts, contributing to the advancement of sustainability assessment methods and governance practices in universities.

Abstract: Drone-view mixed reality (MR) in the Architecture, Engineering, and Construction (AEC) sector faces significant self-localization challenges in low-texture environments, such as bare concrete sites. This study proposes an adaptive sensor fusion framework integrating thermal and visible light (RGB) imagery to enhance tracking robustness for diverse site applications. We introduce the Effective Inlier Count (Neff) as a lightweight gating mechanism to evaluate the spatial quality of feature points and dynamically weight sensor modalities in real-time. By employing a 20 ×16 grid-based spatial filtering algorithm, the system effectively suppresses the influence of geometric burstiness without significant computational overhead on server-side processing. Validation experiments across various real-world scenarios demonstrate that the proposed method maintains high geometric registration accuracy where traditional RGB-only methods fail. In texture-less and specular conditions, the system consistently maintained an average Intersection over Union (IoU) above 0.72, while the baseline suffered from complete tracking loss or significant drift. These results confirm that thermal-RGB integration ensures operational availability and improves long-term stability by mitigating modality-specific noise. This approach offers a reliable solution for various drone-based AEC tasks, particularly in GPS-denied or adverse environments.

Abstract:

B vitamins are essential micronutrients for human health with prominent antioxidant properties, capable of scavenging reactive oxygen species (ROS) and maintaining redox homeostasis, protecting cells from oxidative damage. To address global nutrient deficiencies and identify plant-based antioxidant sources, this study quantified seven B vitamins (B₁, B₂, B₃, B₅, B₆, B₉, B₁₂) in seeds, leaves, and seedlings of five oilseeds (sesame, peanut, soybean, rapeseed, perilla) and two leafy vegetables (spinach, lettuce) via LC-MS/MS, revealing distinct species- and tissue-specific patterns. Notably, sesame seeds exhibited exceptional vitamin B₃ (niacin, 39.3 μg/g), surpassing other oilseeds by 1.6-8.2-fold; its leaves contained outstanding vitamin B₆ (2.88 μg/g), with 2.57–8.31-fold higher than spinach (1.12 μg/g) and lettuce (0.34μg/g), whereas, vitamin B₁₂ (0.44 μg/g), with levels of ~13–20 times higher than other leaves samples. Sesame seedlings recorded high vitamin B₆ (1.6 μg/g) and B₁₂ (0.1 μg/g) among the oilseed crops seedlings. These findings highlight sesame as a multifunctional B vitamin resource for antioxidant nutrition, supporting dietary optimization, crop biofortification, and mitigation of global B vitamin inadequacies via plant-based solutions.

Abstract: Background/Objectives: Repetitive magnetic stimulation (rMS) and static magnetic stimulation (sMS) are currently used as adjunctive therapies for certain neurological conditions. Despite substantial advances in cancer treatment, unfavorable prognoses and outcomes persist, especially for aggressive neoplasms, including glioblastoma and acute myeloid leukemia. In this context, the application of magnetic fields has demonstrated significant anti-tumoral benefits in both in vitro and animal studies, indicating its potential as an effective non-invasive therapeutic strategy; nevertheless, the precise mechanisms of action remain unclear. This scoping review was intended to identify published research investigating the effects of sMS and rMS in in vitro and in vivo models to evaluate their impacts on morphological and molecular parameters. Methods: Four databases (PubMed, Embase, Web of Science, and Scopus) were assessed; the search strategy was limited to the past twenty-five years of data publication. Studies employing rMS or sMS as a primary therapy for conditions apart from neoplasms, and those not addressing these interventions as an adjuvant therapy were excluded. Results: Nine articles using rMS were included: three in vitro, two employing animal models, and the remaining four including both cellular and animal-based analyses. Seventeen studies using sMS were identified: thirteen in vitro and four in vivo. Conclusions: This review indicates that sMS and rMS are employed as adjuvant therapies for increasing the efficacy of conventional drugs like chemotherapy. Their efficacy relies on specific factors: type of cancer, location, cell type, metabolism, and exposure parameters, including intensity, frequency, and duration.

Abstract: Background: Multimorbidity, the presence of two or more chronic health conditions in an individual, presents a significant challenge for healthcare systems worldwide. Physical activity (PA) is an important intervention for the management of chronic health conditions and prevention of disease complications. However, individuals with multimorbidity face unique barriers to PA participation. Artificial intelligence (AI) has emerged as a promising tool to enhance digital health interventions, offering tailored PA promotion. This review synthesised the current evidence on trials using AI-integrated digital intervention tools (including machine learning, natural language processing and predictive analysis) designed to support PA among individuals with multimorbidity.Methods: A rapid review was conducted following PRISMA guidelines. A comprehensive search was performed across six electronic databases (MEDLINE, EMBASE, CINAHL, OVID, Cochrane Library, PsycINFO, Scopus) covering studies from January 2015 to May 2025. Eligible studies were randomised controlled trials (RCTs) involving adults (≥18 years) with multimorbidity using AI-informed digital health interventions to promote PA. Two reviewers independently screened the articles and extracted the data. Owing to the heterogeneity of the included studies, meta-analysis was not possible, and the results were narratively synthesised.Results: Our initial search identified 276 studies. After removing duplicates and screening titles, abstracts, and full texts, 4 studies met the inclusion criteria. All included studies were RCTs that used AI-integrated digital interventions to promote PA in adults with multimorbidity. AI technology interventions included personalised mobile applications (n=2), decision-support systems (n=1), and socially assistive robotics (n=1). The study populations ranged from generically described multimorbid individuals to those with specific cardiometabolic and respiratory combinations of multimorbidity. PA outcomes were assessed through both self-report questionnaires and objective fitness measures. Attrition was common, particularly in longer-duration studies. While some improvements in PA have been reported, overall evidence remains limited and heterogeneous.Conclusions: The limited number of RCTs suggests emerging but inconclusive evidence on the effectiveness of AI-integrated digital health interventions to support PA in multimorbid individuals. Interventions may offer benefits, but heterogeneity in study design, population, and outcomes limits generalisability. Further research using consistent data collection and outcome measures, as well as longer-term follow-up, is needed.

Abstract: Background: Vulvar diseases remain underreported and possibly under-recognised in Nigeria due to limited awareness, primarily, poor health-seeking behaviour, and absence of structured screening programmes. Vulvar self-examination (VSE) has been proposed as a low-cost method for early detection of vulvar pathology. Objective: To assess the knowledge, attitudes and practices surrounding vulvar self-examination and determine vulvar disease prevalence in a community-based Nigerian cohort. Methods: This cross-sectional observational study was conducted in September 2025 across three centres (two urban and one rural). Women attending a community cervical screening programme were recruited through convenience sampling. Participants completed a survey assessing knowledge, attitudes and practices related to VSE. Clinicians performed vulvar examinations, and detailed findings were recorded. Descriptive and inferential statistics were used. Results: A total of 183 women participated, with only 2.2% of women demonstrating some knowledge of structured VSE. Over 95% admitted they had benefited from the VSE education. The prevalence of vulvar disease was 15.8%, with all conditions being benign. Increasing age, urban residence and longer duration of menopause were significantly associated with higher odds of vulvar disease, though not statistically significant. Conclusion: Knowledge and practice of vulvar structured self-examination are poor among Nigerian women and represent a significant unmet need. Structured education on VSE may facilitate earlier detection of vulvar disease and improve outcomes.

Abstract: The rapid evolution toward sixth-generation (6G) wireless networks introduces Integrated Sensing and Communication (ISAC) as a key enabler for intelligent and resource-efficient systems. Traditional resource allocation schemes for ISAC primarily focus on maximizing spectral efficiency, sensing accuracy, or energy efficiency. However, as networks increasingly support semantics-driven applications, the fidelity of transmitted information becomes equally critical. In this paper, we propose a semantic-aware resource allocation mechanism for 6G ISAC systems that leverages the Deep Deterministic Policy Gradient (DDPG) reinforcement learning algorithm. Unlike conventional approaches, our method explicitly incorporates semantic constraints into the optimization process, prioritizing semantic fidelity while jointly enhancing sensing accuracy and energy efficiency. Simulation results, benchmarked against 3GPP’s emerging 6G standards, demonstrate that the proposed mechanism achieves notable performance improvements across all three dimensions, highlighting its potential to support the next generation of intelligent, context-aware communication systems.

Abstract: As remarked by Boltzmann, the Second Law of Thermodynamics is notable for the fact that it is readily proved using elementary statistical arguments, but becomes harder and harder to verify the more precise the microscopic description of a system. In this article, we investigate one particular realization of the 2nd Law, namely Joule heating in a wire under electrical bias. We analyze the production of entropy in an exactly solvable model of a quantum wire wherein the conserved flow of entropy under unitary quantum evolution is taken into account using an exact formula for the entropy current of a system of independent quantum particles. In this exact microscopic description of the quantum dynamics, the entropy production due to Joule heating does not arise automatically. Instead, we show that the expected entropy production is realized in the limit of a large number of local measurements by a series of floating thermoelectric probes along the length of the wire, which inject entropy into the system as a result of the information obtained via their continuous measurements of the system. The decoherence resulting from inelastic processes introduced by the local measurements is essential to the phenomenon of entropy production due to Joule heating, and would be expected to arise due to inelastic scattering in real systems of interacting particles.

Abstract: The Michelson–Morley experiment yielded a null result, indicating equal light travel times in the longitudinal and transverse arms of an interferometer, traditionally interpreted as evidence against a light-propagating medium. This paper re-examines this conclusion by postulating that space itself possesses elastic properties and constitutes the fundamental medium. Beginning with this premise and modeling matter as standing waves within this space-medium, we first demonstrate that the complete mathematical framework of Special Relativity—including Lorentz transformations, time dilation, and mass-energy equivalence—emerges naturally from the Doppler deformation of these wave patterns under motion. We then extend this wave-mechanical approach to gravity, showing that the Newtonian potential and inverse-square law can be interpreted as the gradient of a spatial deformation field, with gravitational interaction energy arising from the overlap of these deformations. We show that Special and General Relativity emerge as effective geometric descriptions of an underlying elastic dynamics of space, in which relativistic effects correspond to physical deformations of wave-based matter. This framework preserves all empirical predictions of relativity while providing a unified mechanical interpretation of inertia, gravitation, Equivalence Principle, and spacetime curvature.

Abstract: Papua New Guinea is one of the least studied regions in the Southwest Pacific, and large areas of the country, such as the Fly Plat-form, remain poorly understood due to limited exposure and access constraints. This study presents the first documentation of basaltic volcanism on the Fly Platform, based on new field discoveries at Mea-hill and Yemsigi, two areas located approximately 25 km apart. Inte-grated field observations, petrography, mineral chemistry, and whole-rock geochemistry show that both basalt suites were derived from a similar magma source but record contrasting emplacement histories. Meahill basalts, which include welded tuffs and highly ve-sicular basalt units, reflect rapid magma ascent, vigorous degassing, and locally explosive activity. In contrast, the massive, less vesicular porphyritic basalts at Yemsigi preserve a quieter emplacement history, but with more extensive post-magmatic alteration. Geochemical sig-natures from least altered rocks of both suites support an intraplate origin with similarities to Pliocene-Pleistocene lava fields of Northeast Queensland. The origin of the intra-plate basaltic magmatism is enig-matic, but both young volcanic provinces correlate spatially with a lower mantle anomaly that may represent residual slab material and a seated-seated magma source. These findings provide further insight into the tectono-magmatic evolution of the Fly Platform region and highlight the need for continued geological investigation in this underexplored district.

Abstract: Large Language Model (LLM)-based agents have fundamentally reshaped artificial intelligence by integrating external tools and planning capabilities. While memory mechanisms have emerged as the architectural cornerstone of these systems, current research remains fragmented, oscillating between operating system engineering and cognitive science. This theoretical divide prevents a unified view of technological synthesis and a coherent evolutionary perspective. To bridge this gap, this survey proposes a novel evolutionary framework for LLM agent memory mechanisms, formalizing the development process into three stages: Storage (trajectory preservation), Reflection (trajectory refinement), and Experience (trajectory abstraction). We first formally define these three stages before analyzing the three core drivers of this evolution: the necessity for long-range consistency, the challenges in dynamic environments, and the ultimate goal of continual learning. Furthermore, we specifically explore two transformative mechanisms in the frontier Experience stage: proactive exploration and cross-trajectory abstraction. By synthesizing these disparate views, this work offers robust design principles and a clear roadmap for the development of next-generation LLM agents.

Abstract: Contemporary infrastructure management confronts unprecedented challenges arising from ageing systems, resource constraints, and escalating demands for efficiency and sustainability. This review examines the transformative potential of integrating smart sensor networks with predictive analytics and machine learning (ML) to address these challenges through data-driven, proactive management approaches. Smart sensors enable continuous, real-time monitoring of critical infrastructure parameters, including structural integrity, environmental conditions, and operational performance, thereby facilitating early detection of anomalies and potential failures. When combined with predictive analytics and ML algorithms—ranging from regression models and decision trees to neural networks and support vector machines—these sensor data streams enable infrastructure managers to transition from reactive maintenance strategies to predictive and preventive paradigms. This paper synthesises evidence from diverse applications across smart cities, structural health monitoring, and energy utilities, demonstrating substantial improvements in operational efficiency, cost reduction, and asset longevity. Case studies illustrate how predictive models optimise traffic flow, enhance grid reliability, detect pipeline leaks, and forecast structural deterioration. Whilst acknowledging persistent challenges related to data quality, system scalability, model interpretability, and cybersecurity, this review highlights the considerable promise of sensor fusion techniques, edge computing, and autonomous systems in advancing infrastructure management practices. The findings underscore that interdisciplinary collaboration and continued technological innovation are essential to realising fully intelligent, adaptive infrastructure networks capable of meeting the complex demands of urbanisation and sustainability in the twenty-first century.

Abstract: Tyrosine kinase inhibitors (TKIs) added to chemotherapy have improved outcomes ofadult patients with Philadelphia-positive B-cell acute lymphoblastic leukemia (Ph+ B-ALL). These improvements initially led to a larger proportion of patients realizing allogeneic stem cell transplantation (alloSCT), long considered essential for cure, but there has been a re-evaluation of alloSCT. At Princess Margaret Hospital (PM), adult patients with Ph+ B-ALL have been treated with a pediatric-inspired chemotherapy protocol with mostly imatinib. In the last two decades, we have witnessed many iterative changes in our approach. Here we examine the outcomes of all Ph+ B-ALL patients treated at our institution from 2001 to 2019. During this time, there were two major protocol changes – omission of asparaginase in 2009, and discontinuation of routine referral for first complete remission (CR1) alloSCT from the early 2010s. Median follow-up was 41.13 months (range, 0.46-228.79). 141 patients (91.56%) achieved CR1. Patient outcomes improved iteratively, with best results seen in the final (2016-2019) cohort: no asparaginase, no routine alloSCT referral in CR1; 4-year OS and RFS were 87.0% and 69.3%, respectively. The long-term OS in this patient group retained statistical significance in the multivariable analysis (p=0.0176) when BCR::ABL1 molecular residual disease (MRD) were considered.

Abstract: Generative Artificial Intelligence (AI) is transforming quality management (QM) and auditing by expanding automation, supporting data-driven decisions, and enabling more personalized stakeholder interaction. However, its adoption also raises concerns related to system robustness, operational resilience, and regulatory compliance, including potential deviations from Critical-to-Quality (CTQ) requirements, gaps in traceability, and misalignment with established quality standards. This paper proposes a structured conceptual framework for proactive, generative AI-enabled QM and auditing, organized into three functional domains: supplier performance, in-process control, and post-market feedback. The framework shows how generative AI can: 1) strengthen supplier oversight via automated documentation and early risk identification; 2) improve in-process control through real-time anomaly detection and Statistical Process Control (SPC)–based triage; and 3) enhance post-market surveillance using predictive analytics for warranty clustering and prioritized Corrective and Preventive Action (CAPA) preparation. To ensure compliance and auditability, the framework incorporates policy-based constraints, human-in-the-loop checkpoints, and end-to-end digital traceability. Verification was performed through a proof-of-concept case study spanning discrete manufacturing and process-based production environments, comparing a conventional quality workflow with a generative AI-augmented alternative. Expert assessment indicated that the generative AI-assisted workflow achieved better performance on key criteria, including documentation completeness, defect detection, process stability, governance and time efficiency. The obtained results suggest that the proposed framework can support a shift from reactive quality control towards predictive and preventive improvement while preserving alignment with quality standards and organizational quality objectives.

Abstract: Background: Major depressive disorder (MDD) is a highly heritable psychiatric condition with complex polygenic architecture. Competing hypotheses emphasize glutamatergic/synaptic plasticity deficits or neurodevelopmental synaptic pruning dysregulation, but integrated testing across large-scale genetic data remains limited.Methods: We re-analyzed the latest Psychiatric Genomics Consortium MDD GWAS (approximately 358,000 cases and 1.28 million controls, European ancestry) using gene-based and competitive gene-set testing (MAGMA), partitioned heritability (LDSC), transcriptome-wide association studies (TWAS with GTEx v8 brain models), and two-sample Mendelian randomization (MR) with cognitive reserve proxies (e.g., educational attainment).Results: MAGMA and LDSC revealed robust enrichment in synaptic pruning-related gene sets (Bonferroni-corrected p < 0.001; up to 1.30-fold LD-adjusted heritability enrichment, p < 10-91), surpassing glutamatergic/plasticity sets (moderate MAGMA enrichment, p = 0.014; no LDSC signal). TWAS showed modest glutamatergic enrichment (1.10-fold mean |Z|, p = 0.007) with heterogeneous directions, while pruning sets were null in TWAS despite strong polygenic signals. MR demonstrated causal protective effects of genetically proxied cognitive reserve on MDD risk (e.g., educational attainment OR = 0.72, 95% CI [0.66–0.79], p = 7.53 × 10-14).Conclusions: These findings prioritize developmental synaptic pruning dysregulation as the primary polygenic substrate of MDD, with downstream impairments in neuroplasticity and cognitive reserve mediating vulnerability. We propose a "pruning-mediated plasticity deficit" framework, integrating neuroimmune and circuit-level mechanisms, with implications for novel therapeutics targeting pruning pathways or plasticity enhancers.