搜索结果：PeerJ

The mode of reproduction is a key aspect for an organism to adapt to its environment. However, understanding the reproductive strategies in extinct animals from deep time remains challenging, due to the scarcity of fossils providing such information. Although viviparity is well-documented across various ichthyosaur lineages, our understanding of the reproductive characteristics of early-diverging Triassic taxa is still limited, particularly regarding their litter sizes. Here, we describe an exceptionally preserved gravid specimen of the Middle Triassic ichthyosaur Mixosaurus from southwestern China, containing at least eleven fetuses. This represents the largest litter sizes ever recorded in a Triassic ichthyosaur, and even one of the highest numbers of fetuses known in all ichthyosaurs. The well-articulated fetal skeletons exhibit consistent developmental stages and advanced ossification, indicating a late prenatal or perinatal state. Their variable body orientations further suggest the absence of a fixed head-first or tail-first birth posture. The presence of over ten fetuses in Mixosaurus demonstrates that high fecundity, potentially indicative of a more r- than K-selection reproductive strategy, had evolved early in ichthyosaur evolutionary history. This reproductive trait may have been a significant factor contributing to the success of this ichthyosaur group with enormous populations and widespread distribution across Northern Hemisphere during the Triassic.

Migration is a well-established risk factor for psychosis, yet limited research has examined long-term clinical trajectories following first-episode psychosis (FEP) in migrant populations within Southern European settings. This study investigates how country of origin and time since migration influence psychiatric outcomes in a naturalistic FEP cohort. A 5-year longitudinal study was conducted on 184 patients with FEP attending the Early Intervention Service from Hospital Universitari de Mataró. Patients were grouped by geographic origin (Spanish, Latin American, Maghrebi, Sub-Saharan) and migration recency (≤5 years vs. >5 years). Primary outcomes included treatment adherence, use of long-acting injectable (LAI) antipsychotics, relapse, and service disengagement. Logistic regression and survival analyses were used to identify associations. Maghrebi origin was associated with lower treatment adherence (OR = 0.32, 95% CI [0.12-0.89], p = 0.030). In time-to-event analyses, Maghrebi patients showed a higher hazard of service disengagement compared with Spanish-born patients (HR = 2.42, 95% CI [1.10-5.34], p = 0. 028). Recent migrants (≤5 years in Spain) also had a significantly increased hazard of disengagement (HR = 2.89, 95% CI [1.30-6.40], p = 0.009). Sub-Saharan origin was strongly associated with earlier initiation of long-acting injectable antipsychotics (HR = 3.40, 95% CI [1.58-7.32], p = 0.002), alongside higher overall odds of LAI use. No significant differences were observed in relapse rates across groups. Both migrant origin and recent arrival were associated with differences in psychiatric care trajectories following FEP. These findings highlight the need to consider adaptation of treatment approaches for migrant populations according to their origin and migration context in order to improve long-term outcomes.

Aerotaxis, the oxygen-directed motile behavior intrinsically linked to bacterial energy metabolism, serves as a fundamental survival and adaptive strategy for motile prokaryotes, including bacteria and archaea. This behavior enables these microorganisms to navigate along oxygen gradients and colonize metabolically optimal ecological niches. This review systematically synthesizes the latest research advancements in bacterial aerotaxis, focusing on the diverse signal-sensing mechanisms of aerotaxis receptors, which are primarily categorized into energy-dependent and direct oxygen-sensing types. Furthermore, we elucidate the distinct molecular mechanisms that underlie their oxygen signal perception and transduction. We also discuss the two major adaptive regulatory mechanisms of aerotaxis and its critical ecological functions in mediating microbe-host interactions, facilitating host colonization, and modulating biofilm formation. This work clarifies the evolutionary and ecological significance of aerotaxis in prokaryotic environmental adaptation and provides a comprehensive theoretical framework for future relevant research.

The yeast Komagataella phaffii (formerly Pichia pastoris) is a robust host for recombinant protein production, capable of growing at high cell densities, performing post-translational modifications, and efficiently secreting heterologous proteins. However, gene expression still relies mainly on the methanol-inducible AOX1 promoter (PAOX1) and the constitutive GAPDH promoter (PGAP), highlighting the need for alternative promoters with distinct regulatory properties. This study aimed to identify and characterize promoters from the most highly transcribed endogenous genes across diverse culture conditions as strong promoter candidates to expand the genetic toolbox for this yeast. Seven promoters were identified through RNA-seq analysis from genes showing transcript levels comparable to or higher than those of genes regulated by PAOX1 or PGAP under specific culture conditions. These promoters were further transcriptionally characterized by reverse transcription-quantitative polymerase chain reaction (RT-qPCR) under different carbon sources, growth phases, and temperatures through the measurement of transcript levels of their corresponding genes, revealing distinct transcriptional profiles. Two major clusters were identified: one associated with genes upregulated during the stationary phase and the other with genes upregulated during the exponential phase. These transcriptional profiles provided a basis for the rational design of fed-batch strategies tailored to each promoter when driving heterologous gene expression. The 0208 and HSP12 promoters showed particular potential at low specific growth rates, favoring high recombinant protein titers with reduced biomass accumulation. In silico analyses suggested that 0208 encodes a mitochondrial membrane-associated protein with a possible functional relationship to Hsp12. Overall, this work provides well-characterized K. phaffii strong promoter candidates, supporting future biotechnological applications.

Deer antler is the only mammalian organ capable of periodic complete regeneration. Its astonishing growth rate-reaching several centimeters per day-provides a unique model for research in tissue regeneration and developmental biology. This rapid growth relies on the protective and signaling functions of the skin, as well as the vigorous proliferation and differentiation capacity of mesenchymal cells. Although previous studies have identified some key factors involved in antler growth, systematically comparing the transcriptomic profiles of these two core tissues (skin and mesenchyme) to identify candidate genes regulating their synergistic growth is crucial for deciphering the molecular mechanisms underlying this "ultra-fast" growth. The Illumina NovaSeq 6000 sequencing platform was used to perform transcriptome sequencing on the antler tissues of the Cervus elaphus xanthopygus (Dongbei red deer) with a growth period of approximately 60 days. Bioinformatics methods were used to identify the transcriptional expression of genes in the antler tissues, and the expression levels of some genes were tested by fluorescence quantitative PCR. A total of 109.82 Gb of transcriptome data was generated from the skin and mesenchymal tissues of antlers, and 40,023 transcripts were obtained. There were 3,530 genes with differential expression between the skin and the mesenchymal tissues of antlers. Genes with upregulated expression levels in the mesenchymal layer were enriched in phosphate-containing compound metabolic processes and multiple biological processes related to bone development such as skeletal system development, cartilage development, regulation of ossification, and enamel mineralization. Among the genes enriched in the cancer pathway and the PI3K-Akt signaling pathway, those that may play an important role in the rapid growth process of deer antlers were screened, including MDM2, RPS6KB1, CASP3, BMP4, PTCH1, SOX-6, YWHAZ, and HSP90B1. The discovery of these genes provides a reference for a further in-depth understanding of the mechanism of the rapid growth of antlers.

Anaerobic power production is not limited to the lower extremity muscles but is also closely related to the stabilization and endurance capacity of the trunk muscles. In particular, the isokinetic properties of the lumbar region muscles may be associated with short-term high-intensity physical performance. The aim of this study was to evaluate the associations between lumbar region isokinetic strength parameters and body composition variables obtained at angular velocities of 60°/s and 240°/s with anaerobic power. In the study, regularized regression analysis (LASSO) with leave-one-out cross-validation (LOOCV) was applied using isokinetic variables such as peak torque (PT), total work (TW), and average power (AP) measured at angular velocities of 60°/s and 240°/s, along with body composition parameters (body fat percentage, body mass index, fat-free mass). Anaerobic power was used as the dependent variable. At 60°/s, isokinetic parameters alone demonstrated limited predictive ability (LOOCV R 2 = 0.19), although several variables such as total work and peak torque were retained in the model. When body composition variables were included, explanatory performance improved substantially (LOOCV R 2 = 0.81), with fat-free mass (FFM) and fat mass emerging as the most stable predictors. At 240°/s, isokinetic parameters alone showed no meaningful predictive ability (LOOCV R 2 = -0.18), indicating poor generalizability. However, when body composition variables were added, model performance increased markedly (LOOCV R 2 = 0.80), with FFM, fat mass, and fat percentage demonstrating consistent associations with anaerobic power. Lumbar isokinetic strength parameters showed limited and unstable associations with anaerobic power when considered independently. However, the inclusion of body composition variables, particularly fat-free mass, substantially improved model performance. These findings suggest that anaerobic power is more strongly associated with body composition than with isolated lumbar isokinetic strength parameters, and that the contribution of trunk muscle function may be secondary and context-dependent.

Terrestrial pollution discharge induces spatiotemporal heterogeneity in riverine nitrogen dynamics. Accurately identifying sensitive temporal phases, spatial hotspots, and primary pollution sources is essential for effective riverine environmental management. This study developed an identification framework to delineate spatiotemporal pollution hotspots and source contributions in the Dai River basin by coupling intensive monitoring with the export coefficient method. The results indicated that the flood season was a critical period for increased total nitrogen (TN) pollution. Approximately 60% of the annual rainfall occurred in July and August, resulting in a 2.5-fold increase in TN concentrations in August compared to June. Moreover, rainfall events exceeding 100 mm/day led to 2- to 3.5-fold increases in TN concentrations relative to pre-event levels, with maximum increases surpassing 14-fold. Automatic monitoring data revealed a decreasing trend in TN concentrations from upstream to downstream. Intensive monitoring identified the Yu, Xidai, and Sha rivers at the headwaters as pollution hotspot zones, characterized by higher nitrogen concentrations and larger basin areas. The export coefficient model results showed that urban domestic wastewater and livestock farming were the dominant sources, accounting for 35.87% and 35.55% of total discharges, respectively. Yuguan town exported 33.34 t of nitrogen, contributing 47% of the total discharges. Based on hotspot identification, we proposed differentiated and integrated management measures for each town.

Metabolic rate assays are essential for assessing organismal stress and resilience, yet their use in aquaculture remains limited. We evaluate a whole-organism, resazurin-based metabolic assay for oysters (Crassostrea gigas and C. virginica) across multiple experimental contexts. Resazurin fluorescence was strongly correlated with oxygen consumption, validating its use as a metabolic proxy. Thermal performance assays revealed expected temperature-dependent metabolic patterns, including clear optima and tipping points where stimulation shifted to metabolic depression. Acute thermal stress experiments showed that individuals exhibiting greater metabolic depression were more likely to survive, indicating metabolism as a predictor of mortality. We also detected genetic variation in metabolic responses, with significant family-level differences. Finally, metabolic rates of 50 selectively bred C. virginica families were significantly correlated with predicted performance. Together, these results establish resazurin as a reliable, scalable assay for whole-organism metabolism, offering a practical tool to support stock selection, hatchery management, and climate-resilient aquaculture.

Nonspecific neck pain (NSNP) is a common musculoskeletal condition that often leads to substantial functional disability and reduced quality of life. Exercise therapy is central to its management. However, the comparative effectiveness of specific modalities and dose-related patterns remains unclear. We conducted a systematic review and network meta-analysis (NMA) including 51 randomized controlled trials with 4,140 participants. Functional disability, measured using validated instruments, was the primary outcome. A Bayesian random-effects NMA was performed to compare exercise modalities, and a dose-response analysis examined the association between exercise parameters and outcomes. Effects were summarized as standardized mean differences (SMDs) with 95% credible intervals (CrIs). Mind-body exercise was associated with a greater reduction in NSNP-related functional disability than control (SMD -0.363; 95% CrI [-0.516 to -0.151]). Resistance training (SMD -0.359; 95% CrI [-0.611 to -0.102]), flexibility training (SMD -0.352; 95% CrI [-0.623 to -0.078]), and coordination training (SMD -0.328; 95% CrI [-0.579 to -0.076]) were also associated with reductions in NSNP-related functional disability relative to control. Exploratory dose-response analysis suggested a possible non-linear association between per-session exercise dose, expressed pragmatically as MET-min, and NSNP-related functional disability. Estimated reductions were observed from around 100 MET-min per session, with larger study-level effects around 300-400 MET-min per session. Mind-body exercise, resistance training, flexibility training, and coordination training appear to be relatively favorable modalities for reducing NSNP-related functional disability. The dose-response findings should be interpreted cautiously, given the heterogeneity of the interventions and the limitations of the available study-level data.

Byzantinia is a genus of extinct cricetodontine rodents with a distribution centered in the eastern Mediterranean (Greece, Turkey, and Lebanon) and extending into Romania. It originated during the Middle Miocene in Anatolia and became extinct at the end of the Miocene, its last representatives being Byzantinia pikermiensis, B. uenayae, and B. hellenicus. Several isolated molars attributed to B. pikermiensis and B. hellenicus have been recovered from a newly discovered fossiliferous locality in Lebanon (Zahleh 16). This material provides new insights into the occlusal morphology of the genus and allows the identification of several evolutionary trends. The co-occurrence of B. pikermiensis and B. hellenicus suggests an age of 8.9-7.5 Ma for Zahleh 16. Given that the earliest record of Byzantinia in Lebanon dates to ca. 10.9-10 Ma, these findings indicate that the presence of the genus in this area persisted for more than 1 million years.

The involvement of oxidative stress (OS) in pre-eclampsia (PE) has been reported, and the present study probed into the OS-related feature genes in PE. The dataset GSE60438 was used to identify the OS-related features in PE, and corresponding feature genes were identified using machine learning algorithms including weighted gene co-expression network analysis (WGCNA), Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis and support vector machine-recursive feature elimination (SVM-RFE). The diagnostic value of the feature genes and their correlation with immune infiltration were explored. Gene set enrichment analysis (GSEA) was performed to identify enriched pathways of these genes. Furthermore, regulatory networks involving the feature genes were constructed. The role of these genes in hypoxia/reoxygenation (H/R)-induced trophoblasts was investigated. Two WGCNA-identified modules (MEpurple and MEturquoise) were intersected with OS-related genes, yielding 359 common genes. Through machine learning algorithms and expression validation, three genes (CPT1A, CYGB and MRPL3) were identified and used to establish a diagnostic model, which reached an AUC of 0.930. The transcription factor (TF) SPI1 and multiple miRNAs formed a potential regulatory network for CPT1A, CYGB and MRPL3, and the expression of these three genes was closely associated with the infiltration of various immune cells (such as monocytes and macrophages M1) in PE. Cellular experiments confirmed that CPT1A silencing alleviated trophoblast dysfunction, reduced reactive oxygen species (ROS) levels, and decreased apoptosis under H/R conditions. The present study examined the OS features in PE, hoping to contribute to the management of PE.

Revitalization measures currently represent an important tool in ecological practice for the recovery of disturbed habitats, and their effectiveness is increasingly assessed using bioindicator groups, including ground beetles (Carabidae). Between 2020 and 2023, we conducted research in the European habitat of the Danube Floodplains, where we evaluated the impact of revitalization measures on ground beetle assemblages. The study was carried out at 15 study sites representing eight habitat types. At each site, five pitfall traps were installed in a linear arrangement. In total, 5,292 individuals belonging to 47 species were analyzed, with a mean ellipsoid biovolume of 192 mm3 per individual. PCA analysis revealed a division of species into two main clusters according to their preference for forest and open habitats, with reference sites without management characterized by a higher proportion of apterous and brachypterous species. In contrast, habitats subjected to restoration measures showed dominance of macropterous species typical of disturbed and dynamic ecosystems. Significant differences in ellipsoid biovolume and all morphometric traits were confirmed among habitats and study years. Habitats with restoration measures exhibited a gradual increase in ellipsoid biovolume and morphometric traits of individuals, indicating improved food availability and more favorable living conditions. Prediction of ellipsoid biovolume development using LSTM models (neural networks) indicated stabilization of Carabidae assemblages in forest habitats after the completion of restoration measures, whereas pronounced fluctuations persisted in open habitats, suggesting the need for a longer time period or more intensive interventions to achieve community stability. The results provide a practical basis for planning and evaluating the effectiveness of restoration and management measures in European habitats of conservation importance, as well as for biodiversity conservation and adaptive landscape management, particularly in optimizing interventions in forest and open ecosystems to achieve long-term stability of Carabidae assemblages.

This study aims to systematically evaluate the impact of exercise interventions on reproductive endocrine profiles in adult patients with polycystic ovary syndrome (PCOS) using a three-level meta-analytical approach. We performed a systematic search of PubMed, Web of Science, The Cochrane Library, and Embase for randomized controlled trials (RCTs) investigating the effects of exercise on reproductive endocrine profiles in PCOS patients, spanning from database inception through April 7, 2026. A three-level meta-analysis was conducted in R using a random-effects model fitted via restricted maximum likelihood (REML). Effect sizes were quantified as Hedges' g with 95% confidence intervals (CIs). Global heterogeneity was assessed using the Cochran's Q test, and the robustness of results was verified through leave-one-out sensitivity analyses. Model fit was compared using likelihood ratio tests (LRT). Publication bias was scrutinized via Egger's test, with the trim-and-fill method employed as a corrective measure where necessary. Methodological quality was appraised using the PEDro scale, and the certainty of evidence was graded via the GRADE profiler. Nineteen studies involving 1,018 participants were synthesized. Exercise interventions significantly lowered testosterone levels (g = -0.34; 95% CI [-0.61 to -0.07]; P = 0.0146). Subgroup analysis revealed that the "aerobic exercise" group was statistically significant (k = 14, g = -0.5120, 95% CI [-0.8230 to -0.2010], P = 0.0027). The mean body mass index (BMI) "25 to 29.9" group was also statistically significant (k = 17, g = -0.4859, 95% CI [-0.7775 to -0.1943], P = 0.0025), as was the mean age group of "26 to 30 years old" (k = 13, g = -0.4878, 95% CI [-0.8465 to -0.1290], P = 0.0105). In conclusion, exercise interventions exert a favorable impact on serum testosterone (T) levels in adult patients with polycystic ovary syndrome (PCOS). PROSPERO (registration no. CRD420251108464; https://www.crd.york.ac.uk/prospero/).

Birds can assess risk posed by predators in their environment and adjust their anti-predation strategies accordingly. However, the impact of nest site characteristics on nest defense behaviors has not been thoroughly studied. This study aimed to investigate the relationship between nest defense behaviors and nest site characteristics in Asian tits (Parus cinereus) when confronted with Common chipmunks (Tamias sibiricus, ground predator) and Eurasian sparrowhawks (Accipiter nisus, aerial predator). Generalized linear mixed-model analyses revealed that predator type significantly influenced both aggressive and vigilance behaviors, with Asian tits less likely to exhibit these two behaviors toward Eurasian sparrowhawks than toward Common chipmunks. Nest-site shrub height negatively predicted the probability of vigilance behavior, whereas no measured nest-site characteristics influenced aggressive behavior. Neither predator type nor nest-site features significantly affected stationary observation or threatening behaviors. These findings demonstrate that Asian tits adopt predator-specific defensive strategies: they adjust vigilance behavior s according to local shrub height while maintaining stationary observation and threatening behaviors relatively invariant, supporting their capacity for dynamic risk assessment in heterogeneous predation environments.

The single leg squat test (SLST) is commonly used for visual movement assessment in subjects with lower extremity injuries. While current methods utilize ordinal rating categorization, adopting a visual analog scale (VAS) may offer potential advantages. The study aims to determine the intra-rater, inter-rater reliability, and validity of a visual analog scale (VAS) for rating the single leg squat test (SLST) in subjects with lower extremity injuries. 58 subjects with lower extremity injuries participated. Intra-rater reliability was assessed by a physical therapist at 1-week and 2-week intervals. Single session inter-rater reliability was assessed by two physical therapists. A 10 cm VAS was used to measure trunk lateral flexion, hip adduction, and dynamic knee valgus during the SLST. Intra-class correlation coefficients (ICC) assessed intra-rater and inter-rater reliability. Validity was assessed by comparing VAS scores to 2-D kinematic analysis. Intra-rater reliability was good at both 1 week (ICC = 0.81-0.92, MDC = 1.2-2.2 cm) and 2 weeks (ICC = 0.77-0.79, MDC = 2.7-3 cm). Inter-rater reliability was good to excellent (ICC = 0.78-0.93, MDD = 1-2.6 cm). The correlation between visual assessment and 2-D kinematic analysis was moderate to good (r = 0.66-0.78). Preliminary data indicate that VAS reliably and validly rates SLST in individuals with lower-extremity injuries, allowing calculation of minimal detectable change and minimal detectable difference to support clinical interpretation. The MDC and MDD values established for trunk lateral flexion, hip adduction, and dynamic knee valgus allows clinicians to objectively determine if scores between repeated measurement and two raters are different.

This study investigated the effects of different velocity loss (VL) thresholds during squat exercise at 80% of one-repetition maximum (1RM) on post-activation performance enhancement (PAPE) in basketball players. Nineteen male collegiate basketball players completed three familiarization sessions and four randomized crossover experimental sessions: control (passive rest), 5% VL, 10% VL, and 20% VL. Countermovement jump height (JH), peak power output (PPO), and reactive strength index modified (RSIm) were measured pre-test and at 3-min and 8-min post-squat. Linear mixed-effects models (LMMs) evaluated the effects of time, condition, and mean-centered relative strength on the outcome variables and repetition counts. To identify responders who exhibited meaningful improvements in jump height, the smallest worthwhile change (SWC) was calculated. At the group level, no significant condition &#xd7; time interactions or main effects of condition were observed for JH, PPO, or RSIm (p > 0.05), indicating no uniform PAPE effect across protocols. Furthermore, no statistically significant strength &#xd7; time &#xd7; condition interaction effects were observed. Repetition counts did not differ between the 5% and 10% VL conditions but were significantly lower in both conditions compared to the 20% VL condition (p < 0.01). Individual analysis showed the 20% VL condition yielded the highest number of responders (n = 11; 58%). Squat protocols using 5-20% VL at 80% 1RM did not produce significant group-level CMJ performance improvements, suggesting that a comprehensive warm-up may create a performance ceiling. Furthermore, these responses were not moderated by relative strength. However, the high inter-individual variability and responder rates at higher VL thresholds highlight the importance of individualized PAPE programming.

With the rising incidence of hematologic malignancies (HMs) in recent years, caregivers of HMs patients have increasingly emerged as a significant factor affecting the treatment of these conditions. This study aimed to explore the association of coping strategies on the relationship between care burden and social support among family caregivers of HMs patients. A questionnaire survey was conducted involving 207 family caregivers of hospitalized HMs patients in a tertiary hospital in Dalian from October 2023 to May 2024 and 200 valid responses were collected. We used hierarchical linear regression and parallel mediation model to assess the relationships and potential mediating effects among care burden, coping strategies, and social support. In our study, the care burden showed a medium level and average score was 44.21&#xa0;&#xb1;&#xa0;15.92. Positive coping strategies (r&#xa0;=&#xa0;&#xa0;-&#xa0;0.769, P&#xa0;<&#xa0;0.01) and social support (r&#xa0;=&#xa0;&#xa0;-&#xa0;0.654, P&#xa0;<&#xa0;0.01) showed a negative correlation with care burden, while negative coping strategies was significantly correlated with care burden (r&#xa0;=&#xa0;0.788, P&#xa0;<&#xa0;0.01). According to the mediating effect analysis after covariate adjustment, the relationships between social support and care burden were mediated by coping strategies in total samples and the indirect effect value was -0.955, accounted for 73.57% of the total effect. Family caregivers of HMs patients with more negative coping strategies and less social support were significantly associated with higher levels of care burden. Coping strategies served as a partial parallel mediator between social support and care burden.

Spatial transcriptomics provides unprecedented insights into gene expression in the spatial context of tissues; however, some mainstream techniques lack quantitative analysis of uncertainty. In this study, we propose a computational framework-Spatial Deconvolution via Deep Gaussian Processes (SDDGP)-that integrates deep Gaussian processes with Bayesian inference for spatially aware deconvolution of spatial transcriptomics data. The framework consists of three tightly coupled components. A multi-layer deep Gaussian process (DGP) captures nonlinear spatial dependencies in cell-type composition by propagating spatial coordinates through successive Gaussian process (GP) layers, each equipped with sparse variational inducing points and a radial basis function (RBF) kernel that can be combined with an external Matern spatial kernel. The DGP output serves as an adaptive, spatially informed prior for per-spot cell-type proportions, which are then estimated via a Negative-Binomial Markov chain Monte Carlo (MCMC) sampler with automatic proposal tuning, thinning, and split-chain convergence diagnostics. We compared our method with existing approaches on two datasets: a pancreatic ductal adenocarcinoma (PDAC) dataset with matched single-cell RNA sequencing and spatial transcriptomics, and a human thymus dataset. Across both datasets, our method estimated cell type proportions more accurately than existing methods, particularly for rare cell types, and recovered biologically meaningful spatial patterns such as the distribution of T cells and plasma cells in PDAC. The Bayesian formulation also provides a quantification of uncertainty for each estimate, which aids the interpretation of spatial heterogeneity in complex tissues.

Deep vein thrombosis (DVT) is a common thrombotic condition with substantial morbidity when not identified early. Machine learning (ML)-based predictive models may improve early identification of patients at high risk for DVT, but few clinically applicable early-risk models exist. To develop and internally validate a ML model using routinely available clinical and laboratory indicators for early risk prediction of DVT, and to identify the most influential predictors using model explainability techniques. We retrospectively analyzed clinical data from 231 patients evaluated at the Fifth Affiliated Hospital of Southern Medical University between January 2017 and June 2024. Patients were labeled as DVT occurrence (n&#xa0;=&#xa0;159) or non-occurrence (n&#xa0;=&#xa0;72). Seven candidate predictors were selected by Least Absolute Shrinkage and Selection Operator (LASSO) regression. The dataset was split into training (70%, n&#xa0;=&#xa0;162) and test (30%, n&#xa0;=&#xa0;69) sets. Five ML algorithms were trained: XGBoost, CatBoost, Random Forest (RF), Logistic Regression, and Support Vector Machine, with hyperparameter tuning on the training set. Model performance was assessed by 5-fold cross-validation and on the held-out test set using Area Under the Receiver Operating Characteristic Curve (AUC), accuracy, recall, and F1 score. The best model was further interpreted via feature importance and Shapley Additive Explanations (SHAP). LASSO selected seven predictors: hemoglobin, platelet count, leukocyte count, fibrinogen, prothrombin time, D-dimer (DD), and glucose. The Random Forest model showed the best discrimination (test-set AUC = 0.874), with favorable accuracy, recall, and F1 compared with other classifiers (detailed metrics reported in the manuscript). In the RF model, D-dimer had the highest feature-importance contribution; SHAP analysis confirmed DD as the dominant risk driver and characterized the directions and relative effects of other features. We developed an internally validated ML model for early DVT risk prediction using seven routine clinical variables; Random Forest achieved the best performance and identified D-dimer as the most influential predictor. This model may support earlier identification and intervention for patients at risk of DVT, pending external validation and prospective evaluation.