Search Results (1,367)

Background/Objectives: Iron deficiency remains a prevalent condition, needing specific laboratory tests for diagnosis. This study aimed to evaluate whether routine complete blood cell count (CBC) parameters can be used within a machine learning framework to predict low ferritin and low transferrin saturation, used as biochemical markers of altered iron status, potentially supporting more targeted laboratory test utilization. Methods: In this single-center retrospective outpatient study, we analyzed 32,437 records from subjects undergoing both complete blood cell count and iron metabolism testing between 2023 and 2026. Low ferritin and low transferrin saturation were defined using sex-specific thresholds. Low ferritin was present in 14,344 subjects (44.2%), whereas low transferrin saturation was present in 7791 subjects (24.0%). After cleaning data and excluding incomplete records, demographic variables and CBC indices were tested as potential predictors. The dataset was split into training and test sets with stratified sampling. Multiple supervised machine learning models, including logistic regression, decision tree, random forest, XGBoost, support vector machine, k-nearest neighbors, and Naive Bayes, were trained. Hyperparameter tuning and model selection were performed using repeated stratified 10-fold cross-validation, optimizing the area under the curve (AUC). Model performance was assessed by AUC, sensitivity, and specificity, and validated on an independent test set. Results: All models showed predictive capability for low ferritin and low transferrin saturation using CBC parameters alone. Ensemble methods, especially random forest and XGBoost, reached the best performance (AUC values of 0.80–0.87 for ferritin and 0.85–0.96 for transferrin saturation). Sensitivity and specificity were balanced, supporting clinical screening applicability. Results were maintained across validation and confirmed in the test set. Prediction of transferrin saturation showed slightly higher accuracy than ferritin. Feature importance analysis identified mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), and red blood cell distribution width (RDW) as key predictors. Conclusions: CBC-based machine learning models may help identify subjects with low ferritin or low transferrin saturation, supporting subsequent targeted assessment of iron status. Full article

The vertebrate mitochondrial genome (mitogenome) is a small, circular DNA molecule typically ~16–17 kb in length, encoding 37 genes that are essential for the electron transport chain, the mechanism that drives mostly all the ATP synthesis in cells. Owing to its central role in energy metabolism, its structure is highly conserved across vertebrate lineages in both the number and relative position of each gene in the genome. Nevertheless, different variations have been found in several teleost lineages, including antarctic fishes (Nototheniidae), gadiforms, hatchetfishes (Sternoptychidae), and Batrachoidiformes. The explanation for these phenomena remains unknown yet may reflect shifts in functional constraints and can provide insights into lineage-specific and/or coevolutionary processes. This raises the possibility that mitogenome structure is related to habitat selection, potentially reflecting environmental influences on energetic regulation. To further test this hypothesis, we studied more than 400 teleost species across all major teleost lineages. The mitogenome sequences were downloaded from NCBI and annotated using two independent algorithms (MITOZ and MITOS) and then compared with a reference (Danio rerio) to find any deviation from the standard structure. Similarly, ecological data was downloaded from FishBase using the R Package “rfishbase” 5.0.3. Two independent ancestral reconstruction analyses were carried out for both traits, “Mitogenome” and “Habitat”, using a reference evolutionary tree for teleosts to unravel both evolutionary histories. The possible association between mitogenome and habitat was then assessed using a suite of phylogenetic comparative methods, including Pagel’s correlation test (corHMM) to evaluate whether both traits evolved in a correlated fashion, branch-level co-transition analysis to identify lineages where structural changes and habitat shifts co-occurred, and node-by-node comparisons of ancestral state probabilities across the phylogeny. Preliminary results suggest a correlation between some deep-sea environments and a modified mitogenome structure, with structural deviations tending to cluster in lineages inhabiting greater depths. These exploratory findings raise the possibility that changes in mitogenome architecture may be linked to adaptations in energetic metabolism required for life in extreme low-energy environments. Further analyses are underway to clarify the functional significance of these genomic changes and their relationship to ecological and metabolic pressures in teleost evolution. Full article

To address the limitations of the traditional rapidly exploring random tree (RRT) algorithm, including redundant exploration, limited adaptability in dense environments, insufficient obstacle clearance, and poor path smoothness, this paper proposes an integrated multi-strategy improved RRT framework for robot path planning. The proposed method combines KD-Tree accelerated nearest-neighbor search, adaptive step-size adjustment, safety-boundary-based collision checking, direct goal-connection, safety-validated shortcut pruning, and spline-based path smoothing. Comparative experiments were conducted in four representative scenarios, including simple sparse, complex dense, narrow-passage, and highly cluttered environments. Traditional RRT, RRT*, RRT-Connect, and the proposed method were evaluated through 30 independent runs in each scenario. The results show that all algorithms achieved a 100% success rate in the tested scenarios. RRT* generated near-optimal paths but required substantially longer planning time and a much larger search tree, while RRT-Connect achieved the shortest planning time but provided relatively small obstacle clearance. In contrast, the proposed method achieved a better balance among path length, tree compactness, obstacle clearance, geometric smoothness, and computational efficiency. Ablation experiments further verified the contribution of each module, and parameter sensitivity analysis demonstrated the influence of the balancing factor, safety margin, and spline sampling number. The proposed framework provides a practical and lightweight path-planning solution for static two-dimensional environments. Full article

Introduction: Freshwater ecosystems are global biodiversity hotspots, yet they remain highly vulnerable to biological invasions. Non-native freshwater fish species (NNFS) have established self-sustaining populations across nearly all biogeographic realms, reshaping regional ichthyofaunas and driving community-level impacts through predation, competition, hybridisation and ecosystem disruption. Critically, both foreign introductions and within-country translocations (extralimital species) contribute to this process, yet the latter remain more weakly regulated and consistently under-studied in invasion risk frameworks. Objective: We developed a stage-structured profiling framework to jointly evaluate foreign and extralimital NNFS in Greece and predict three sequential invasion outcomes, establishment, spread and integration, with the goal of identifying the ecological traits and pathway variables that best explain invasion success at each stage and informing management policy. Methodology: We compiled a dataset of 63 NNFS recorded in Greek freshwaters (36 foreign, 27 extralimital), characterised by eleven ecological, biogeographic and anthropogenic attributes. Logistic and multiple regression models and classification and regression trees (CART) were fitted independently for each invasion stage, with cross-validated predictor screening to limit multicollinearity and a taxonomy-based covariate to account for phylogenetic non-independence. Results: All 27 extralimital translocations established successfully, compared with only 11 of 36 foreign introductions, underscoring the disproportionate establishment success of within-country movements. Establishment probability was positively associated with high physiological tolerance and proximity to the nearest native source, and negatively associated with maximum adult size; propagule pressure provided only weak additional support. Spread across drainage basins was driven primarily by introduction effort and physiological tolerance. Integration increased with introduction effort, while the CART identified distance from the nearest native source as the primary partition of widespread, high-abundance outcomes, with trophic level further structuring outcomes among extralimital taxa. Conclusions: Our results indicate that management frameworks focused solely on foreign NNFS substantially underestimate invasion risk from within-country translocations. A compact set of predictors, biogeographic proximity, physiological tolerance and introduction effort, offers a practical, pathway-inclusive screening tool to guide prevention, surveillance and early detection in Mediterranean river networks, addressing a recognised European policy gap where extralimital movements remain more weakly regulated than foreign introductions. Full article

Niche and interspecific association are important components of community ecology and are of great significance for revealing the mechanisms of community assembly and its stability. In this study, the woody plant communities of Betula platyphylla Sukaczev forests in the Niyang River Basin of southeastern Qinghai–Tibet Plateau were taken as the research object. The niche, interspecific association, and community stability of dominant tree species in B. platyphylla forests were analyzed using the Levins index (B_L), Shannon index (B_S), Pianka index (O_ik), Schoener index (C_ik), variance ratio (VR), chi-square test, association coefficient (AC), Spearman rank correlation, and M. Godron stability methods. The results showed that a total of 71 woody plant species were recorded across 48 plots, mainly belonging to Rosaceae, Ericaceae, and Caprifoliaceae. B. platyphylla, Quercus aquifolioides Rehder & E. H. Wilson, Sorbus rehderiana Koehne, and Berberis gyalaica Ahrendt had relatively large niche breadths, indicating strong resource utilization ability and a wide range of spatial adaptation. They were the main constructive species and dominant species of B. platyphylla forest communities in this basin. The overall niche overlap of woody plant communities was relatively low, indicating relatively obvious differentiation in resource utilization among different species. Interspecific association analysis showed that the dominant species in the tree layer exhibited an overall significantly positive association, whereas those in the shrub layer exhibited an overall non-significantly positive association. The associations between species pairs were mainly non-significant, and the overall interspecific association was weak. Most species showed a relatively independent distribution pattern, reflecting weak interspecific competition within the community. Community stability analysis showed that the Euclidean distance between the tree layer and the theoretical stability point (20, 80) was 20.17, whereas that of the shrub layer was 27.98, indicating that the tree layer was more stable than the shrub layer. Overall, the community may not yet have reached a fully stable state. The results provide important references for biodiversity conservation, vegetation restoration, and sustainable forest management in alpine canyon ecosystems. Future studies should incorporate environmental factors such as soil properties and hydrothermal conditions to further reveal the ecological mechanisms driving community succession and stability. Full article

Mitochondrial genomes are widely used in insect taxonomy and phylogenetics, but their signals may conflict with morphology and nuclear genomic evidence because the mitochondrial genome represents a single maternally inherited locus. Here, we assembled complete mitochondrial genomes of four pygmy grasshoppers, Zhengitettix transpicula, Formosatettix sp., Gibbotettix parvipulvillus, and Bolivaritettix sp., using PacBio HiFi reads. The four mitogenomes ranged from 15,152 to 17,976 bp and contained the typical 37 mitochondrial genes. Mitochondrial phylogenies inferred by maximum likelihood and Bayesian methods were topologically identical and recovered several well-supported tetrigid relationships, including a close relationship between Formosatettix sp. and Bolivaritettix sp. However, Z. transpicula was unexpectedly placed near Macromotettixoides rather than close to other Zhengitettix representatives. In contrast, a morphology-based tree recovered Z. transpicula with Z. triangularis, and comparison with a published nuclear single-copy ortholog tree based on 1962 loci supported a non-mitochondrial placement of Zhengitettix inconsistent with the anomalous mitochondrial position of Z. transpicula. Independent assembly from the original HiFi reads, read-depth inspection, protein-coding gene checks, and nuclear-genome screening for NUMT-like sequences supported the authenticity of the assembled Z. transpicula mitogenome. These results document mito–nuclear and cyto-morphological discordance in Tetrigidae and highlight the need for integrative interpretation of mitochondrial phylogenies in taxonomically complex insect groups. Full article

Crack opening and reinforcement stress are two complementary indicators of the service state of reinforced concrete hydraulic structures, yet they are often predicted separately. This study develops a data-driven multi-task temporal fusion framework for joint 48 h ahead prediction of dam crack responses and rebar stress using multi-source monitoring data. The measured data comprise five crack-monitoring series, five rebar stress series, local temperature channels, reservoir water level, antecedent rainfall, and an auxiliary environmental signal over approximately four years. Target responses are aligned only at common measured timestamps; no synthetic target observations are introduced. A simplified engineering layout and plan-based crack–rebar distances are further used to examine whether an explicit spatial prior can strengthen the shared temporal representation without introducing synthetic target values. A residual multi-task temporal fusion network (MTTF-Net) is proposed with a shared Transformer encoder, attention pooling, task-specific decoders, and a response-continuity regularization term. The model is compared with persistence, Ridge regression, random forest, Extra Trees, XGBoost, and GRU baselines under a chronological train/validation/test split. For the independent test period, Ridge regression obtains the lowest overall RMSE (2.2968), whereas MTTF-Net provides the lowest crack RMSE (0.0141), the lowest overall MAE (1.0035), and the second-best overall RMSE (2.3813). Distance-informed ablation, denoted as MTTF-Net-S, remains close to MTTF-Net in macro-averaged $R^2$ but is not superior in the overall test metrics, indicating that the available horizontal distances are valuable engineering metadata but cannot replace richer three-dimensional structural connectivity. These results indicate that the monitoring data contain a strong linear autoregressive component, while multi-task temporal fusion improves nonlinear crack response prediction and remains competitive for stress forecasting. The source code is prepared as a public implementation package, whereas the measured monitoring dataset is subject to data owner restrictions. Full article

Background/Objectives: Type 2 diabetes mellitus (T2DM) is a prevalent metabolic disorder associated with substantial long-term morbidity and mortality. Routinely collected anthropometric, biochemical, and hematological variables may contain useful discriminatory information for data-driven classification. This study aimed to compare the apparent classification performance of multiple machine learning algorithms for distinguishing individuals with and without T2DM using routinely obtained clinical parameters in a single-center dataset. Methods: This single-center observational study included 160 adults (95 females, 65 males) evaluated at the Endocrinology Outpatient Clinic of Gaziantep Islam Science and Technology University, Faculty of Medicine, Ersin Arslan Training and Research Hospital. The dataset comprised anthropometric measurements, biochemical markers, and complete blood count parameters. SMOTE was applied only within the training folds to address class imbalance and to avoid information leakage. Following fold-internal data preprocessing, which included imputing missing values and feature standardization where appropriate, the dataset was evaluated using stratified 5-fold cross-validation. SHAP analysis was performed to interpret the model predictions. A calibration curve was used to assess the model’s reliability. Eight supervised machine learning models were evaluated with and without HbA1c: Logistic Regression, Linear Discriminant Analysis, Quadratic Discriminant Analysis, Decision Tree, Random Forest, Extra Trees, Gaussian Naive Bayes, and k-Nearest Neighbors. Model performance was evaluated using accuracy, sensitivity, specificity, and F1 score, and ROC curves were used as a diagnostic tool. Results: The models were evaluated in two different ways: with and without HbA1c. Random Forest demonstrated the best classification performance in the cross-validated evaluation; without HbA1c, it achieved 92.2% accuracy, 93.9% sensitivity, 97.9% specificity, and a 95.9% F1 score. When HbA1c was included, it achieved 98.0% accuracy, 97.9% sensitivity, 98.8% specificity, and a 99.0% F1 score. Decision Tree and Extra Trees demonstrated strong performance with accuracy rates of 87.6% and 92.8%, respectively, without HbA1c, and 90% and 93.5% when HbA1c was included; in contrast, KNN yielded the lowest accuracy rate (70.6%). Overall, tree-based models performed better than linear classifiers on this dataset. Conclusions: Machine learning models based on routine clinical and anthropometric variables demonstrated promising performance for T2DM classification in this single-center dataset; tree-based approaches yielded the most promising results. Including HbA1c improved the models’ ability to classify individuals with and without T2DM. However, since HbA1c was included both as a predictor and as part of the operational definition of the diabetes group, the findings should be interpreted with caution due to the risk of target leakage. Therefore, these results should be considered exploratory rather than evidence of clinically applicable predictive performance, and an independent external validation study should be conducted prior to clinical application. Full article

Human activity recognition—HAR—plays a crucial role in the lives of patients battling neurodegenerative diseases. These debilitating conditions, such as Alzheimer’s or Parkinson’s, affect individuals’ ability to perform daily tasks autonomously and safely. HAR technology offers an invaluable solution by enabling real-time monitoring and assistance, helping to maintain independence and quality of life for patients. Additionally, this technology provides a valuable data source for doctors and caregivers, allowing for more precise and personalized care, which can make a difference in managing and treating these neurodegenerative diseases. The objective of this review is to identify the contribution of Transfer Learning and Reinforcement Learning in supporting the processes of daily activity recognition, thus enhancing the quality of life for patients. As this is a trending topic, the literature surrounding it is quite dispersed, which is why this review aims to present the current line of research in this field. To carry out this analysis, the science tree paradigm was used, which establishes two fundamental stages of analysis. The first is delimited by scientometrics, where the leading countries in the application of such technologies can be identified. This review highlights the evolution in the use of transfer learning and reinforcement learning in HAR in the healthcare field, where these techniques have significantly improved the accuracy and adaptability of real-time monitoring systems. The studies reviewed indicate that transfer learning has allowed models to adapt to data variations without requiring large volumes of manual labeling, which is essential in clinical and patient monitoring contexts. Additionally, reinforcement learning has optimized decision-making in complex scenarios, enabling activity recognition systems to dynamically adjust monitoring parameters, enhancing detection and response to critical or unusual activities in multi-user environments. These advances demonstrate that, by integrating these approaches, greater personalization and robustness can be achieved in human activity recognition, thereby improving the quality of life for patients in clinical settings. Full article

Microbial classification and taxonomic information are fundamental to microbiological studies. Raman microspectroscopy, a rapid and non-destructive single-cell analytical technique, captures intrinsic molecular fingerprints reflecting cellular biochemical composition, thereby enabling microbial classification at the single-cell level. However, current Raman-based classification frameworks allow accurate identification only for micro-organisms already represented in reference databases. These approaches often fail or yield errors for uncharacterized microorganisms. To address this limitation, we collected 6600 single-cell Raman spectra from 11 microbial species, including bacteria and fungi, and developed deep learning models for rapid classification. A hierarchical clustering (HC) framework based on Raman features extracted by a one-dimensional convolutional neural network (1D-CNN) was constructed and compared with phylogenetic trees derived from rRNA gene sequences. 1D-CNN achieved high classification performance with an overall accuracy of 99.7%. Notably, the Raman HC tree exhibited clear concordance with phylogenetic structures, particularly at the higher taxonomic levels. Validation using five independent unknown strains demonstrated that the Raman HC tree consistently positioned these strains near their closest phylogenetic relatives, in strong agreement with sequence-based analyses. Collectively, these findings highlight the potential of single-cell Raman spectroscopy with deep learning as an alternative and complementary framework for microbial taxonomic analysis, particularly for previously uncharacterized microorganisms. Full article

Background: Sport-related injuries are common, and often recurrent, among female college soccer players. This exploratory cohort study investigated whether perceptual decision efficiency and sleep quality could discriminate between injured and uninjured players. Methods: Twenty-seven NCAA Division I women’s soccer players (19.5 ± 1.3 years) completed a perceptual response training program, administered through an immersive virtual reality system, across a 13-week season. Players completed 11 training sessions progressing through four levels of task difficulty, with conjugate eye movements, neck rotation, and whole-body lunge-reach responses measured for each trial. Four metrics, elapsed time, rate correct per second, across-trials variability, and an efficiency index, were calculated for each of three defined time segments: perceptual decision, action initiation, and perceptual–motor response. The Pittsburgh Sleep Quality Index (PSQI) and Global Well-Being Index (GWBI) were administered prior to the first practice session, and all subsequent time-loss injuries were documented. Receiver operating characteristic analyses, Kaplan–Meier survival analysis, and classification tree modeling were used to evaluate injury discrimination. Results: Twelve time-loss injuries, including five concussions and seven lower extremity musculoskeletal injuries, were sustained by 10 of the 27 players. Optimal discrimination between injured and uninjured players was derived from the perceptual decision efficiency (PDE) metric for the most difficult perceptual response training task (AUC = 0.682–0.794), with a binary cut point of ≤6.02 yielding an odds ratio of 5.60 (95% CI: 1.02, 30.90; Mantel–Cox log rank p = 0.025). All five concussions occurred in players classified as high-risk by a suboptimal PDE value. Pre-participation PSQI demonstrated an AUC of 0.735. Notably, no player with both an optimal PDE value and a favorable sleep quality score (PSQI < 4) sustained a time-loss injury. Moderate-to-large training-related improvements in perceptual decision metrics were observed for the least challenging task from early- to late-season sessions. Conclusions: Optimal values for PDE and sleep quality together characterized female college soccer players who avoided injury. Both factors appear to be modifiable, suggesting that perceptual response training combined with interventions to enhance sleep quality may enhance injury resistance. Independent validation in larger, diverse athlete cohorts is warranted. Full article

Phosphate waste rock piles (PWRPs) generated by open-pit phosphate mining are highly heterogeneous and difficult to characterize using conventional point sampling alone, which limits representative resource assessment, selective recovery, and rehabilitation planning. This study develops an integrated framework combining PRISMA spaceborne hyperspectral imagery, ground-based mineralogical analyses, and spatially constrained machine learning to map lithological heterogeneity at the Benguerir phosphate mining site, Morocco. A three-stage spectral optimization workflow, including atmospheric band masking, Savitzky–Golay filtering, and analysis of variance (ANOVA)-based feature selection, was applied to identify the most discriminative Short-Wave Infrared (SWIR) bands for lithological classification. After removing redundant observations located within shared PRISMA pixel footprints, 127 spatially independent samples were retained for model development. Five supervised classifiers (Random Forest, Extra Trees, XGBoost, Support Vector Machine, and K-Nearest Neighbors) were evaluated under a spatially constrained cross-validation framework aligned with the 30 m native PRISMA pixel size. Ensemble-based models, especially Extra Trees and Random Forest, provided the most stable performance, with balanced accuracies of 0.56–0.69 and area under the receiver operating characteristic curve (AUC) values exceeding 0.95 for carbonate-dominated lithologies. Lower discrimination between phosphate and siliceous facies reflects intrinsic mineralogical mixing and spectral overlap at the sensor scale. Entropy-based uncertainty and posterior probability mapping revealed spatially structured prediction ambiguity concentrated along lithological boundaries and transitional zones, consistent with petrographic evidence of compositional heterogeneity. These results indicate that moderate but stable accuracies likely represent realistic performance limits for spaceborne hyperspectral mapping of complex mining environments under spatial constraints. The proposed framework provides a transferable and uncertainty-aware basis for lithological mapping, selective recovery assessment, and sustainable phosphate waste management. Full article

Microwave curing offers energy-efficient geopolymer synthesis, yet optimizing compressive strength remains challenging due to complex variable interactions. This study develops an interpretable stacking ensemble surrogate model to predict the strength of microwave-cured geopolymers. A literature-derived dataset of 235 observations was systematically compiled from literature (2010–2025), covering diverse aluminosilicate precursors, activator concentrations, and curing parameters. The framework integrates linear, kernel, and tree-based base models via a linear meta-learner to enhance generalization. Unlike conventional models, source-aware validation was implemented to ensure reliability across heterogeneous studies. The stacking ensemble significantly outperformed standalone models, achieving a coefficient of determination (R²) of 0.957 and a low RMSE of 5.64 MPa. Crucially, the model demonstrated high reliability through rigorous residual analysis and noise sensitivity stress tests, confirming stable performance across the entire strength range. Interpretability analyses using SHAP and partial dependence plots identified curing time, microwave power, and specimen size as dominant factors governing strength. These dependencies adhered to established geopolymerization kinetics, yielding physically reasonable response surfaces. This work demonstrates that stacking ensemble models serve as reliable statistical surrogate tools (not mechanistic simulators) for preliminary experimental screening within the investigated parameter space. The framework has not been validated against genuinely independent experimental campaigns or industrial-scale microwave curing conditions, and should be treated as a literature-based surrogate model rather than a deployment-validated predictive tool. Full article

Climate change directly affects agricultural productivity, particularly in small island systems where ecosystems and livelihoods are highly exposed to climate variability. This study presents a comprehensive analysis of climate variability for the three districts North and Middle Andaman, South Andaman, and Nicobar, using a six-model CMIP6 ensemble under four SSP scenarios (SSP126, SSP245, SSP370, and SSP585), coupled with ensemble tree-based machine learning algorithms to project coconut yield responses. The historical data was analysed from 1981 to 2025 and the projection was from 2026 to 2100. Observed rainfall reveals a persistent north-to-south gradient, with South Andaman recording the highest mean annual rainfall (3408.40 mm) and Nicobar recording the lowest (2442.13 mm), alongside pronounced inter-annual variability and a discernible drying tendency post-2015. Nicobar consistently records the warmest mean Tmax (30.89 °C) and Tmin (24.11 °C), while North and Middle Andaman exhibit the greatest inter-annual temperature variability. Future projections indicate a robust and statistically significant warming across all districts and scenarios, with end-of-century Tmax increases reaching up to 4.05 °C (Nicobar, SSP585) and Tmin increases up to 3.73 °C (North and Middle Andaman, SSP585), accompanied by a progressive compression of the diurnal temperature range. Precipitation projections show modest wetting in the Andaman districts under most scenarios, while Nicobar exhibits a muted response, with SSP370 uniquely projecting a decline of approximately 69 mm below the observed baseline. Among the ten evaluated CMIP6 models, six (ACCESS-CM2, CMCC-ESM2, CNRM-ESM2-1, EC-Earth3-Veg-LR, GFDL-ESM4, and NorESM2-MM) were selected based on composite skill scores across rainfall, Tmax, and Tmin. Model selection was optimized independently for each district via Leave-One-Year-Out cross-validation with hyperparameter tuning, yielding district-specific best performers: GradientBoost for North and Middle Andaman (R² = 0.471), RandomForest for South Andaman (R² = 0.609), and ExtraTrees for Nicobar (R² = 0.289). K-Nearest Neighbours demonstrated competitive predictive skill in all three districts, confirming that instance-based learning can capture non-linear climate–yield relationships, though tree-based ensembles were preferred for their robustness and interpretability. Ensemble tree-based ML models and instance-based learning consistently outperformed all linear and kernel-based approaches, confirming the non-linear nature of climate–yield relationships in this setting. Coconut yield projections indicate above-baseline productivity gains of 3.4–21.5% in North and Middle Andaman and 24.6–36.8% in South Andaman, driven by favourable warming and precipitation trends, while Nicobar yields plateau at 7.7–13.7% above baseline, indicating thermal saturation of the climate yield response under already near-optimal thermal conditions. Notably, Nicobar exhibits a reversed yield–emission relationship wherein lower-emission pathways marginally outperform high-emission scenarios, likely reflecting avoidance of thermal stress thresholds. Inter-CMIP6-model uncertainty emerges as the dominant source of projection spread, exceeding scenario uncertainty across most districts, underscoring the critical importance of multi-model ensemble frameworks for robust agricultural climate impact assessments in data-sparse tropical island environments. Full article

Accurate viral protein function annotation is essential for genomic surveillance, yet conventional retrieval-augmented generation (RAG) pipelines often fragment biological evidence into fixed-length text chunks, disrupting relationships among ORFs, annotations, structural domains, sequence motifs, residue mappings, and model-derived attention evidence. We propose ViroBioTree, a tree-structured biological evidence retrieval framework for downstream viral protein evidence review rather than a new primary annotation classifier. Built as an evidence organization layer on ViralMultiNet-derived ORF-level predictions and annotations, ViroBioTree converts sequence, annotation, structure, and attention evidence into typed biological nodes and traceable edges, then performs deterministic multi-channel recall, evidence-aware reranking, balanced TopK selection, rule-based verification, and node-cited report generation. In a demo benchmark, ViroBioTree achieved its strongest deterministic proxy performance on structure-explanation tasks, with Precision@K = 1.0, Recall@K = 1.0, and diversity = 0.52; these values reflect expected node-type and tag agreement rather than independent biological correctness. A bounded full-scale SARS-CoV-2 index contained 39,800 ORF rows, 80,000 attention records, 199,418 nodes, and 495,886 edges. In a stratified full20k diagnostic evaluation, ViroBioTree showed task-dependent advantages over LlamaIndex vector retrieval for conflict detection, evidence retrieval, and structure explanation, while LlamaIndex remained competitive or stronger for annotation-rich function annotation. A cross-family Influenza A Virus (IAV) diagnostic audit showed that the schema can represent IAV evidence namespaces while explicitly exposing missing formal ORF inputs, missing attention evidence, and unavailable residue/PDB assertions. Supplementary robustness, external sanity-check, diversity-risk, expert-evaluation, domain-tool positioning, and cross-family audit analyses supported traceability, report quality, and conservative evidence handling, but also showed that stable Precision@K under query perturbation does not necessarily imply stable retrieved evidence sets. ViroBioTree operates offline and deterministically, but does not address raw-read assembly, base calling, primary ORF prediction, or wet-lab validation. Its results should be interpreted as proxy and expert-reviewed evidence for traceable viral protein evidence retrieval and report generation rather than as direct validation of biological function annotation. Full article