Despite the fast progress of deep learning, one standing challenge is the gap of the observed training samples and the underlying true distribution. There are multiple reasons for the causing of this gap e.g. sampling bias, noise etc. In the era of foundation models, we show that when leveraging the off-the-shelf (vision) foundation models (e.g., CLIP, DINOv2) for feature extraction, the geometric shapes of the resulting feature distributions exhibit remarkable transferability across domains and datasets. To verify its practical usefulness, we embody our geometric knowledge-guided distribution calibration framework in two popular and challenging settings: federated learning and long-tailed recognition. In the federated setting, we devise a technique of acquiring the global geometric shape under privacy constraints, then leverage this knowledge to generate new samples for clients, in the aim of bridging the gap between local and global observations. In long-tailed learning, it utilizes the geometric knowledge transferred from sample-rich categories to recover the true distribution for sample-scarce tail classes. Comprehensive experiments show that our proposed geometric knowledge-guided distribution calibration effectively overcomes information deficits caused by data heterogeneity and sample imbalance, with boosted performance across benchmarks.

Observational studies often rely on nonprobability samples that may not represent the target population, limiting the generalizability of findings. Health examination data from the Korean National Health Insurance Service (NHIS) faces this challenge, as voluntary participation introduces sampling bias. Poststratification offers a potential solution by reweighting samples to match population distributions. Using the NHIS-National Sample Cohort, we demonstrated three poststratification approaches-simple poststratification, raking, and multilevel regression with poststratification-to estimate obesity prevalence among adults 20-39 years of age in 2019. The Korea National Health and Nutrition Examination Survey served as the reference standard for population-level estimates. We compared these methods with inverse probability sampling weights. Additionally, we applied poststratification to evaluate the accuracy of self-reported disease history for hypertension, diabetes, dyslipidemia, and stroke. Crude obesity prevalence from NHIS-National Sample Cohort was 36.3% (95% confidence interval: 36.0, 36.6), substantially higher than the Korea National Health and Nutrition Examination Survey reference of 31.4% (28.8, 33.9). Simple poststratification using age and sex reduced this estimate to 33.9% (33.6, 34.2), ranking with additional smoking and alcohol variables yielded 34.7% (33.9, 35.5), and multilevel regression with poststratification incorporating region produced 34.2% (33.3, 35.0). Inverse probability sampling weights yielded comparable results (33.9%-34.1%). For self-reported disease history, poststratification consistently produced modest decreases in sensitivity estimates, suggesting that health examination participants may report disease history more accurately than the general population. Poststratification provides a principled framework for improving population-level inferences from nonprobability samples. These methods warrant broader application in epidemiological research using administrative and electronic health record data.

Prognostic relevance and molecular correlates of Claudin-1 expression in pancreatic neuroendocrine tumors.

Abstract Background Trabectedin, a marine-derived antineoplastic agent, has shown promise in various malignancies, prompting investigation into its efficacy for high-grade meningiomas. Objective This systematic review aims to evaluate the efficacy and safety of trabectedin in the management of high-grade meningiomas by synthesizing data from existing randomized controlled trials (RCTs). Methods Comprehensive searches were conducted in PubMed, the Cochrane Database of Systematic Reviews (CDSR), and Epistemonikos using MeSH terms for “Trabectedin” and “Meningioma.” Studies were included if they focused on trabectedin treatment in meningioma patients and were RCTs. Preclinical studies and trial protocols were excluded. Data extraction included study design, patient characteristics, intervention details, outcomes, and adverse events. Quality assessment was conducted using the Cochrane Risk of Bias tool. Results we got three RCTs, with one meeting the inclusion criteria. This RCT, conducted by Preusser et al., involved 90 patients with recurrent WHO grade 2 or 3 meningiomas. The trabectedin arm showed an overall survival of 13.54 months compared to 10.61 months in the local standard of care (LSOC) arm. However, progression-free survival (PFS) was shorter in the trabectedin arm (2.43 months) compared to the LSOC arm (4.17 months). The quality assessment revealed a high risk of bias due to the open-label design. Conclusion Trabectedin demonstrates a modest survival benefit in patients with high-grade meningiomas. The high risk of bias and limited sample size underscore the need for larger, double-blind RCTs to validate these findings. Further research should explore combination therapies and identify biomarkers to optimize treatment efficacy and patient selection.

A gated integrating laser-combined scanning tunneling microscope has been developed for high-fidelity investigation of photophysical processes at the atomic scale. This instrument integrates a low-repetition-rate, high-pulse-energy laser with a synchronized gated-integration scheme, enabling selective extraction of the laser-induced tunneling current. To demonstrate its performance, atomically resolved topography and surface photovoltage mapping were simultaneously obtained on the Si(111)-(7 × 7) surface. The surface photovoltage distribution is found to be strongly correlated with the underlying atomic structure. In addition, a highly localized enhancement of the laser-induced tunneling current is observed at a specific single-atom defect site. Detailed analysis, supported by local density of states measurements, reveals that this defect possesses an unusually high density of deep valence-band states. This distinctive electronic structure likely promotes Auger recombination between photogenerated holes accumulated under positive sample bias and electrons tunneling from the tip, thereby giving rise to the observed increase in the local tunneling current.

Estimating the prevalence of genetic disorders is complicated by many factors including sampling bias and differing methods of estimation. However, establishing the true prevalence of these disorders is critical for understanding disease burden, pharmacoeconomic modeling, and resource allocation for testing and care. Phelan-McDermid syndrome (PMS) is a genetic neurodevelopmental disorder with an unknown true prevalence and previous estimates vary from 2.5-10 per million births. The study team reached out to a multitude of sources, including clinical genetic testing laboratories, research centers, and clinical centers. Sites provided the number of PMS diagnoses made out of the total number participants with autism tested at their site. Further extrapolations were made to adjust for the proportion of individuals with PMS who do not have autism, autism diagnosis age limitations, and type of genetic variant. Lastly, Centers for Disease Control estimates of autism rates were used to extrapolate to the general population. Ten sources participated and data from 179,837 autism cases were evaluated. The frequency of PMS diagnoses ranged from 1% to ~2.5%. However, the studies had different levels of sensitivity depending on the assay(s) used, and therefore raw results should not be directly compared. After applying extrapolations, including adjustments for assay sensitivity and other factors, the final weighted average was 13.7 per 100,000 (95% CI 10.02-18.60 per 100,000), indicating 1 in ~7300 individuals in the general population have PMS. We leveraged a diverse set of genetic data from multiple sources to generate a population-level estimate of the prevalence of PMS. Our findings indicate that PMS affects approximately 13.7 per 100,000 individuals, substantially higher than previous estimates.

This study is the first to examine awareness, knowledge, and attitudes towards dyslexia among the general public in Mainland China. Using an online survey, we collected data on (a) demographics, (b) awareness of dyslexia and other common neurodevelopmental conditions, (c) knowledge of dyslexia causes, symptoms, functional impact, and assessment/intervention, and (d) attitudes towards dyslexia. A total of 1,008 adults from across all major regions of China completed the suvey. Around 70% reported having heard of dyslexia, lower than awareness of autism and ADHD, but higher than that of developmental language disorder. Respondents answered 49% of knowledge items correctly and demonstrated greater knowledge of dyslexia symptoms, followed by functional impact and causes, with weaker knowledge of dyslexia assessment and intervention. Dyslexia awareness and knowledge were higher among younger adults, females, urban residents, non-parents, and those with higher education and income, with some variation across regions. Attitudes towards dyslexia were generally positive, following similar demographic patterns. Although greater awareness was associated with higher levels of knowledge, only dyslexia knowledge uniquely predicted attitudes towards dyslexia after controlling for demographic factors. These indings highlight the need for culturally relevant, awareness-raising campaigns that promote a more accurate understanding of dyslexia. The findings should be interpreted in light of the limitations linked to sampling bias and methods of data collection. Future studies should include the voices of individuals with dyslexia to better understand how social and cultural factors in China influence their lived experiences across development.

Deep learning of CT imaging predicts PD-L1 expression and immunotherapy response in metastatic NSCLC: A multi-center study.

Next Generation Sequencing (NGS) based microbial surveillance has become a critical tool for public health initiatives, providing insights into complex interactions within microbial communities. This understanding can be used to facilitate outbreak investigations, pathogen surveillance, and resistance prevention. However, current sample preparation methods are prone to human error and sample bias resulting in skewed Gram +/- coverages and low yields. Methods that have been optimized for non-bias extraction, like mechanical bead beating, are not easily integrated with automation, preventing widespread scalability. Here, we present an innovative, fully automated method that integrates enzymatic lysis, extraction, and library preparation in a single-operator, single-cartridge workflow. The workflow was designed to substantially reduce hands-on time and minimize Gram-stain-related biased extraction. A capillary-based liquid handler is leveraged to reduce hands-on time from 8 to 10h to less than 45min. Using mixed microbial communities, we demonstrated that the workflow consistently produces high-quality sequencing libraries with an average yield of 80.5 ng/µL, average quality score of 33.6, and using the ZymoBIOMICS Microbial Community Standard, a 2.46-fold improvement in Gram-positive representation relative to a standard lysis protocol. To our knowledge, this is the first demonstration of a fully integrated capillary-based workflow combining enzymatic lysis, extraction, and library preparation compatible with low-throughput platforms. While automated NGS preparation systems exist, they predominantly rely on high-throughput liquid handling robotics, require separate equipment/workflows for extraction and library preparation, or do not support unified workflows for Gram-positive and Gram-negative extraction. The present approach addresses the low-throughput, cost-sensitive, rapid-turnaround requirements of public health and clinical microbiology settings. This proof-of-concept study demonstrates a fully automated workflow for microbial NGS sample preparation, benchmarked on a standardized mock community. This work offers a transformational engineering foundation with the potential to increase efficiency and reproducibility of pathogen genomics workflows, enabling faster outbreak detection and more comprehensive surveillance programs.

Ticks are haematophagous ectoparasites and vectors of a diverse range of pathogens, exerting substantial impacts on wildlife, domestic animals and public health. In Australia, despite the country's rich and unique biodiversity, a comprehensive understanding of ticks and tick-borne pathogens associated with wildlife remains limited. Environmental change, urban expansion and climate variability are increasingly disrupting wildlife habitats, potentially intensifying interactions between wildlife hosts, ticks and humans. A broad evidence synthesis of studies published between January 1940 and March 2024 was conducted, retrieving 133 eligible records from Web of Science, CABI Abstracts and PubMed databases. Fifty tick species parasitising 160 wildlife species were identified, predominantly from the genera Ixodes, Amblyomma and Haemaphysalis. The most commonly reported hosts included marsupials, particularly bandicoots, wallabies and possums, with notable tick species being Ixodes tasmani, Ixodes holocyclus and Amblyomma triguttatum. Microorganism records were relatively limited and mostly represented molecular detections or reported associations, including Babesia, Borrelia, Coxiella, Rickettsia and Theileria species, rather than confirmed vector competence, reservoir status or pathogenicity. Key limitations included geographic sampling biases towards eastern Australia, limited molecular identification of ticks and infrequent pathogen screening, particularly regarding the ecology, epidemiology and molecular diversity of host-vector-microorganism interactions. Improved surveillance, expanded molecular characterisation, and integrated One Health investigations are required to better understand the ecological and public health significance of these host-vector-microorganism interactions.

Intensive care unit (ICU) patients are at high risk for pressure injuries (PIs), medical device-related pressure injuries (MDRPIs), and skin tears (STs). Although nurses play a key role in preventing all three conditions, existing research has mainly focused on PIs, with limited evidence addressing nurses' combined knowledge related to MDRPIs and STs. In particular, shared deficiencies across these conditions remain unclear. This study aimed to assess ICU nurses' knowledge of PIs, MDRPIs, and STs together to identify overlapping knowledge gaps that may affect skin integrity management in ICU settings. A cross-sectional online survey was conducted with 283 ICU nurses. Data were collected using three validated tools: the Pressure Ulcer Knowledge Assessment Tool 2.0 (PUKAT 2.0), the Medical Device-Related Pressure Injury Knowledge Assessment Tool (MDRPI-KAT), and the Skin Tear Knowledge Assessment Instrument (OASES), enabling construct-specific measurement across related conditions. Knowledge levels varied across domains, with mean total scores of 14.44 ± 3.43 (PUKAT 2.0), 8.73 ± 1.83 (MDRPI-KAT), and 10.17 ± 2.44 (OASES). Across all three tools, prevention consistently emerged as the lowest-scoring domain. Higher knowledge scores were observed among nurses with greater professional experience, postgraduate education, and prior care experience related to STs. The findings provide a comprehensive overview of ICU nurses' knowledge regarding PIs, MDRPIs, and STs, highlighting variability across domains and consistently lower scores in prevention-related areas. These results underscore the need for targeted educational strategies to strengthen knowledge across key aspects of skin integrity management in intensive care settings. However, findings should be interpreted with caution due to potential sampling bias related to social media-based recruitment and the absence of multivariable analysis. Not applicable.

Conventional two-dimensional (2D) pathology relies on a limited number of tissue sections and therefore provides information from isolated planes, which can introduce sampling bias and lead to the loss of axial details, making it difficult to accurately capture the full structural organization of tumors and their microenvironment. In recent years, the integration of three-dimensional (3D) imaging technologies, such as tissue clearing and light-sheet fluorescence microscopy, combined with virtual staining, artificial intelligence, and spatial omics, has enabled the acquisition and analysis of volumetric biological information while preserving the overall architecture of the tissue. Following the theme of "advancing from planar observation to 3D quantitative analysis", this paper outlines the major technical pathways of 3D pathology, from sample clearing and fluorescent labeling to 3D imaging and visualization and then to an analytical framework for 3D pathological data. Based on this, we further examined representative applications, including tumor staging, lesion detection, and immune microenvironment profiling, to illustrate the advantages and potential clinical value of 3D pathology over conventional 2D approaches in precision diagnosis, prognostic assessment, and therapeutic decision-making. We also highlight key challenges that remain for 3D pathology, such as storage and computational costs, workflow and diagnostic standardization, and clinical implementation. Overall, AI-enabled 3D pathological imaging holds promise as an important direction for pathology, although its integration into routine clinical decision-making will require continued technological development and translational research.

Liquid biopsy in gynecologic cancers: Cellular origin to early detection and precision care.

BackgroundAccess to geriatric mental health (GMH) care is limited in rural areas. To meet this need, the Veterans Health Administration provides specialty tele-GMH care for aging rural veterans via regional telehealth hubs.ObjectiveThis study aims to create a roadmap describing key phases and determinants underlying the implementation and sustainment of tele-GMH services as part of a qualitative longitudinal evaluation of tele-GMH teams.MethodsSemistructured interviews were conducted with clinicians from all 8 tele-GMH teams (n=25) at 3 time points across a 3-year period (October 2021-September 2024). Interview (n=46) data were summarized into key domains using a templated rapid qualitative approach, guided by the Consolidated Framework for Implementation Research (CFIR) 2.0. Further thematic analysis and team discussion elucidated the findings.ResultsWe identified key activities and determinants of success in three phases: (1) preimplementation (engaging leaders, securing funding/hiring, and defining services); (2) implementation scale-up and expansion (advertising, addressing challenges, seeking feedback, refining, and growth); and (3) sustainment (maintenance). Activities within each phase were cyclical and iterative (ie, nonlinear). Barriers to implementation included unfamiliarity with local aging resources; facilitators included tailoring strategies and engaging referring clinicians.ConclusionsSimilar processes emerged across regions in the development and sustainment of tele-GMH services, allowing for the creation of a unified roadmap. Limitations including sampling bias are discussed. Further work could apply and evaluate the utility of the roadmap to guide creation of tele-GMH services in new regions to enhance access to specialty care for aging rural veterans.

This systematic review integrates interdisciplinary literature on the impact of women's leadership on firm performance, based on 40 core articles (AJG 2024 4*/4) and 10 supplementary landmark studies. Four theoretical perspectives are identified: social roles, resource supplementation, contextual factors, and behavioral practices. The review reveals that the influence of women's leadership is not a simple positive or negative effect, but a complex process dynamically regulated by bias challenges, resource complementarity, situational conditions, and strategic behaviors. Core debates center on the shift from simple judgments to situational dependence, from stereotype constraints to strategic integration of agency and communion, and from numerical representation to qualitative capital. Key gaps are identified, including the lack of multi-level interaction theories, sample biases toward developed countries, and endogeneity issues in causal identification. Future research should focus on multi-level interactive processes and expand exploration in Global South contexts such as China.

The concept of ichnodisparity is herein introduced for the study of fossil vertebrates, using Gondwanan avian footprints. We define ichnodisparity for trackways as the qualitative variation in morphological patterns of autopodial impressions, most likely reflecting functional or ecological thresholds. Foot architectural designs are defined based on anatomical features, including digit number and orientation and interdigital webbing, and size. Ichnodisparity in fossil bird tracks is proposed here as a useful proxy for analyzing evolutionary radiations and characterizing macroevolutionary patterns. Twenty-three geological formations from Gondwana were analyzed, spanning from the Early Cretaceous to the Late Pleistocene. The earliest morphotypes were predominantly large-sized, while small morphotypes became dominant over time. No zygodactyl footprints are recorded from Gondwana. Prior to the K/Pg boundary, ichnodisparity was high, consistent with a Cretaceous avian radiation. During the Eocene, morphological variety remained comparable to that of the Late Cretaceous - except for the large tridactyl design - suggesting ecological turnover rather than functional expansion; elevated ichnodisparity may be associated with the Early Eocene Climatic Optimum. The decline recorded in the Oligocene is likely attributable to sampling bias. The Miocene concentrates the highest ichnodisparity in the fossil record, with the reappearance of large pedal architectures linked to cursorial birds, supporting an expansion into open spaces. In the Pleistocene, small tridactyl forms disappeared and the large palmate architecture emerged for the first time, indicating a new episode of ecological turnover.

Background:Acinetobacter baumannii is a major nosocomial pathogen with extensive antimicrobial resistance, yet the extent to which resistome burden covaries with mobilome architecture across clonal backgrounds remains unclear at the global scale. Using a curated dataset of 10,500 high-quality genomes, this study examined lineage-resolved relationships among AMRFinder-detected resistance determinants, insertion sequence (IS) burden, plasmid diversity, predicted plasmid mobility, and virulence-associated gene presence. Results: The population structure was highly uneven and dominated by a small number of lineages. Using total AMRFinder burden, genome-level associations with IS burden and plasmid replicon diversity were modest (Spearman rho = 0.259 and 0.246, respectively). In a sensitivity analysis using an acquired-like burden metric that excluded OXA-51 family hits, blaADC family hits, predefined intrinsic or conserved chromosomal background loci, and non-antibiotic resistance classes, the median burden decreased from 17 to 9 hits per genome. Under this filter, acquired-like burden remained associated with IS/mobile genetic element burden (rho = 0.237) and showed a slightly stronger association with plasmid replicon diversity (rho = 0.285), with both relationships persisting after excluding genomes from China and the United States. Mobilome features remained markedly heterogeneous across lineages, whereas virulence-associated gene presence profiles were comparatively conserved. Conclusions: These findings support a lineage-structured view of resistance-associated genomic variation in A. baumannii. However, the observed genome-wide associations were small in magnitude and should be interpreted as patterns of co-occurrence rather than direct evidence of mechanism or causality. Sensitivity analyses suggest that plasmid-related features show a more robust relationship with acquired-like resistance burden than IS counts alone. This distinction is relevant for genomic surveillance, but it should be interpreted together with lineage imbalance, sampling bias, and the limitations of gene-presence-based inference.

ENSO-driven distribution shifts in the critically endangered Mustelus whitneyi along the Northern Peruvian Coast.

Free energy perturbation (FEP) methods have become indispensable for predicting relative binding free energies (RBFE) in lead optimization, yet they often suffer from slow convergence due to high-energy barriers in ligand conformational sampling. In this work, we introduce and validate a terminal-flip Monte Carlo (TFMC) move specifically designed to accelerate conformational sampling in alchemical simulations. We integrated TFMC into alchemical RBFE calculations of protein-ligand complexes and benchmarked it on JNK1, cMet, SHP2, Thrombin, and Factor Xa systems. TFMC yields a 2-5× faster convergence of key dihedral distributions and improves ΔΔG estimates by up to 1 kcal/mol compared with conventional molecular dynamics (MD) simulation-based FEP. Our method removes sampling biases in scaffold-hopping transformations and is straightforward to implement in existing FEP workflows.

Emergency coronary artery bypass grafting (CABG) in high-risk patients remains a major perioperative challenge, and surgical strategy is often determined by clinical and anatomical factors rather than randomization. This study aimed to describe early outcomes of emergency CABG within a structured institutional technique selection framework. This retrospective cohort study included 59 consecutive patients who underwent emergency isolated CABG between 2004 and 2007. Patients were allocated to on-pump (n = 39) or off-pump (n = 20) CABG according to a predefined institutional algorithm incorporating hemodynamic status, coronary anatomy, and comorbidities. Outcomes are presented descriptively, and between-group comparisons are interpreted in the context of inherent selection bias. Substantial baseline and procedural differences were observed between groups, reflecting structured technique selection rather than random allocation. Patients undergoing on-pump CABG had more complex coronary anatomy and higher clinical acuity, including all four cardiogenic shock cases and 92.6% of triple-vessel grafting procedures. Longer operative time, increased inotropic support, and longer hospital stay were observed in the on-pump group; CK-MB levels at 1h were lower in the off-pump group. In-hospital mortality was low and comparable (7.7% vs. 5.0%, p = 1.000). Given the marked selection bias and limited sample size, these findings should be considered descriptive rather than comparative. In this historical cohort of emergency CABG patients, both on-pump and off-pump techniques were feasible within a structured institutional selection framework. Observed differences in early outcomes likely reflect underlying patient characteristics rather than intrinsic superiority of either technique. These findings highlight the importance of individualized, algorithm-guided decision-making in emergency surgical revascularization.