ABSTRACT Myopia is a major global public health challenge, characterized by excessive axial elongation and an increased risk of irreversible vision‐threatening complications. In clinical practice, management remains experience‐based and is constrained by a limited ability to integrate high‐dimensional non‐linear biometric information, creating an unmet need for more precise and individualized decision‐making. Although artificial intelligence (AI) has demonstrated strong performance in screening, diagnosis, and prediction of myopia onset and progression, most existing studies have focused on forecasting disease trajectories rather than generating actionable support for treatment selection and optimization. In this review, we summarize recent advances in AI applications across the continuum of myopia care, from prediction of onset and progression to early assessment of pathological evolution, and further examine emerging AI‐enabled strategies for intervention planning in orthokeratology, pharmacologic therapy, and refractive surgery, with particular emphasis on treatment responsiveness. Finally, we discuss the major barriers to clinical translation, including data heterogeneity, limited interpretability, and implementation costs, and outline a future framework for an integrated patient‐centered AI ecosystem to support precision management and reduce the global burden of myopia.
ABSTRACT Dermatology lacks a centralized analysis‐ready transcriptomic resource, leaving publicly available datasets fragmented and difficult to reuse without bioinformatics expertise. To address this gap, we developed SkinDB, an open‐access code‐free resource that brings together 220 curated human microarray datasets comprising 11,283 samples across six major skin diseases: systemic lupus erythematosus, atopic dermatitis, scleroderma, psoriasis, dermatomyositis, and vitiligo. Five analytical modules provide 18 functions spanning differential expression, expression regulation, pathway analysis, machine learning, and gene‐set analysis, with downloadable figures and result tables. The platform supports both dataset‐specific exploration and recurrence‐based cross‐dataset summaries while retaining cohort context. In a psoriasis demonstration, MKI67 was highest in lesional skin and correlated with CDC20 in GSE13355. Across nine psoriasis datasets, MKI67‐high samples showed recurrent transcription‐factor changes; across all 18 psoriasis datasets, consensus analysis identified 14 pathways associated with high MKI67 expression, led by cell cycle, DNA replication, and DNA repair. By converting dispersed dermatological transcriptomes into an accessible analytical resource, SkinDB supports cross‐cohort exploration, biomarker research, and hypothesis generation.
ABSTRACT Mendelian randomization (MR) is a method that utilizes genetic variants as instrumental variables to determine causal relationships between exposures and outcomes, thereby mitigating confounding bias and reverse causality inherent in observational studies. Rooted in Mendel's laws of inheritance, MR has undergone rapid development since Katan first introduced the concept in 1986, followed by the formalization of its methodology by George Davey Smith and Ebrahim in 2003. This review comprehensively summarizes the theoretical foundations, methodological innovations, and expanding applications of MR. This review discusses the three core assumptions underpinning MR methodologies—relevance, independence, and exclusion restriction—and examines advanced methodological extensions, two‐sample MR, multivariable MR, bidirectional MR, non‐linear MR, cis ‐MR, mediation MR, multi‐population MR, and cluster MR. It further provides an overview of key commonly used R packages, databases, and analytical workflows that facilitate the implementation of MR. Application domains spanning gene–environment interactions research, public health, complex diseases, drug target validation, and integration with other cutting‐edge technologies are highlighted through representative case studies demonstrating their translational potential. Lastly, we critically assess methodological limitations, including weak instrument bias, horizontal pleiotropy, population stratification, data quality heterogeneity, lack of benchmark validation and comparison across MR methods, as well as selection and collider bias, while proposing future directions for improving robustness and expanding the applicability of MR through integration with multi‐omics and cross‐ancestry analyses. Overall, MR serves as a cornerstone in modern causal inference research, bridging genetics, epidemiology, and precision medicine to advance our understanding of disease etiology and therapeutic innovation.
ABSTRACT Gastric cancer remains a health burden, and its metabolic drivers are poorly defined. N‐(2‐furoyl) glycine, a circulating metabolite, has been implicated in tumor biology. We investigated its causal role in gastric cancer and the underlying mechanism. Mendelian randomization using FinnGen and UK Biobank GWAS data evaluated the causal association between N‐(2‐furoyl) glycine and gastric cancer. Single‐cell RNA sequencing of gastric lesions characterized receptor expression, and siRNA knockdown tested dependence on GLRX3. TCGA bulk transcriptomes underwent WGCNA to define a GLRX3‐associated gene module, which was used to train a deep‐learning survival model in TCGA and validate it in external cohorts. The model‐derived risk score was combined with clinicopathologic variables to build a clinical prediction tool. The metabolite enhanced proliferation, clonogenicity, and invasion while suppressing apoptosis. Single‐cell profiling reveals its upregulation in epithelial and endothelial cells of early gastric cancer. GLRX3 knockdown abrogated N‐(2‐furoyl) glycine–mediated tumor promotion. The GLRX3‐centered module supported a deep‐learning survival model that provided consistent risk stratification and improved prognostic performance when integrated with clinical variables. N‐(2‐furoyl) glycine acts as a metabolic driver of gastric cancer via GLRX3. A GLRX3‐based deep‐learning model enhances risk prediction and highlights this axis as a therapeutic target.
ABSTRACT Clear cell renal cell carcinoma (ccRCC) exhibits notable genetic and epigenetic heterogeneity. Although H4K20me3 is markedly enriched in ccRCC samples, the precise functional role of this histone modification in ccRCC pathogenesis remains elusive. The expression and prognostic value of KMT5C, the catalytic enzyme responsible for H4K20me3, were investigated using public databases and clinical specimens from our institute. Functional experiments were conducted to elucidate the role of KMT5C in ccRCC progression. Mechanistically, comprehensive molecular approaches, including peptide pull‐down, shotgun proteomics, co‐immunoprecipitation, microscale thermophoresis, RNA‐seq, ChIP‐qPCR, and RIP‐qPCR, were employed to dissect the signaling pathway. The levels of H4K20me3 and KMT5C were significantly elevated in ccRCC. And elevated level of KMT5C was associated with the poor prognosis. Functional experiments demonstrated the oncogenic role of KMT5C in ccRCC proliferation and migration. Mechanistically, we identified EWSR1, an RNA/DNA‐binding protein, as a direct interaction partner of H4K20me3. Downregulation of EWSR1 significantly inhibited cell proliferation and migration of ccRCC cells. Transcriptomic analysis combined with ChIP and RIP experiments revealed that KMT5C and EWSR1 co‐regulate the expression of ACADM. Specifically, KMT5C‐mediated H4K20me3 directly suppressed ACADM transcription, whereas EWSR1 further reduced ACADM mRNA stability by modulating its m 6 A modification level. Importantly, the combination of A‐196 (KMT5C inhibitor) and sunitinib effectively inhibited tumor growth in a xenograft model, demonstrating significant therapeutic potential. Our study revealed that KMT5C‐mediated H4K20me3 promotes ccRCC progression by suppressing ACADM transcription and destabilizing its mRNA through recruitment of EWSR1. Targeting KMT5C with A‐196 in combination with sunitinib represents a promising therapeutic strategy for ccRCC.
ABSTRACT Musculoskeletal disorders such as osteoarthritis (OA), rheumatoid arthritis (RA), and osteoporosis (OP) affect millions worldwide, often progressing silently until advanced stages. Accessible early risk stratification remains challenging because conventional diagnosis often depends on imaging, specialist evaluation, and fragmented clinical information. Herein, we developed Luma AI, a multimodal AI platform that integrates routine blood tests and demographic variables with curated patient‐reported symptom narratives. The text branch uses a domain‐specific clinical text encoder based on LoRA‐adapter‐fine‐tuned Bio_ClinicalBERT, whereas the structured branch uses optimized classical machine learning models. The two branches are fused through an explicit prediction‐score and text‐representation pipeline for early screening of OA, RA, and OP. Performance was validated using the U.S. NHANES and multi‐center Chinese cohorts. We evaluated more than 50 predictive models and applied Bayesian optimization to refine the structured models, achieving over 80% accuracy with structured data alone and exceeding 90% in the multimodal validation cohort when symptom narratives were included. The platform incorporates explainable AI tools to identify key risk factors and provides personalized post‐screening recommendations for lifestyle, diet, and medical follow‐up. This study demonstrates that Luma AI can support scalable low‐cost early risk screening and personalized management of OA, RA, and OP while remaining positioned as a risk‐stratification aid rather than a stand‐alone diagnostic replacement.
ABSTRACT The Global Burden of Disease (GBD) database provides essential primary data for public health research. However, its complexity and diversity pose challenges for conducting advanced secondary analyses. To address these challenges, we developed easyGBDR , an innovative integrated R toolkit designed to facilitate the secondary analysis of GBD data. This package offers a comprehensive analytical workflow that encompasses data acquisition and preprocessing through to advanced statistical modeling and visualization. easyGBDR supports various epidemiological methods, including trend analysis, age‐period‐cohort (APC) modeling, future disease burden projection, decomposition analysis, frontier analysis, and health inequality assessment. By automating data handling processes and standardizing procedures while generating publication‐ready tables and figures, easyGBDR significantly reduces the technical barriers associated with robust GBD‐based research. This toolkit enhances reproducibility, accelerates evidence generation, and empowers researchers across all levels of expertise to extract meaningful public health insights from one of the world's most comprehensive health databases. Consequently, easyGBDR serves as a vital resource for advancing evidence‐informed policy and practice in global health.
ABSTRACT Helicobacter pylori ( H. pylori ) infections are known to cluster within households. To improve the efficiency of population‐wide H. pylori screening, we therefore developed and validated a questionnaire‐based model to identify households with multiple infections. This model was developed in 2021 based on 10,735 households (31,098 individuals) enrolled from 29 of 31 provinces in mainland China. In 2025, 724 households (2094 individuals) were prospectively enrolled from 3 provinces in China to validate the model. High‐risk households were defined as those containing at least two infected members. After feature selection, we compared 10 machine‐learning classifiers and interpreted the final model using Shapley Additive Explanations (SHAP). The screening efficiency of two different strategies, the screen‐and‐treat strategy (screen all) and the family strategy (screen members in high‐risk households while excluding low‐risk households), was compared in a micro‐simulation of a city of 1,000,000 residents and in an empirical evaluation using real‐world data from 29 provinces. XGBoost achieved the optimal discrimination with an area under the curve (AUC) of 0.861 in the training cohort and 0.839 in the external validation cohort, with corresponding sensitivities of 0.798 and 0.813. In the micro‐simulation, the family strategy identified 276,384 infections with 394,913 tests, whereas the traditional screen‐and‐treat strategy required approximately 680,000 tests to achieve a comparable yield. In the real‐world dataset validation, the family strategy improved detection efficiency from 40.7% to 70.0%, with consistently larger gains in lower‐prevalence provinces. By using a questionnaire‐based model, the family‐based strategy significantly improved screening efficiency and precision in the evaluated datasets, suggesting its potential as a practical tool for household risk stratification and a more efficient H. pylori screening. Further validation is needed in broader geographic and epidemiological settings before wider implementation.
ABSTRACT The relative importance and biological mechanisms of modifiable risk factors for cardiovascular diseases remain unclear among adults with hypertension. In a large prospective cohort of 134,417 adults with hypertension, we evaluated the individual and joint effects of modifiable risk factors on the incidence of eight cardiovascular diseases using Cox proportional hazards models, with joint effects quantified using a composite cardiovascular health score. Population attributable fractions were estimated to quantify and prioritize each risk factor's contribution to cardiovascular disease burden. Mediation analyses quantified the extent to which proteins and metabolites accounted for the associations, and functional enrichment analysis identified underlying biological pathways. During a median follow‐up of 13.26 years, 29,099 participants (21.6%) developed cardiovascular diseases. More favorable risk factor profiles were consistently associated with lower risks across all outcomes, with hazard ratios per 10‐point increase in the cardiovascular health score ranging from 0.92 (95% CI: 0.87–0.97) for transient ischemic attack to 0.76 (95% CI: 0.72–0.79) for aneurysm. Among individual factors, nicotine exposure showed the largest population attributable fractions across multiple diseases (7.9%–36.4%), followed by body mass index. Mediation analyses identified 212 proteins and 111 metabolites that statistically accounted for part of these associations, whereas enrichment analyses suggested involvement of pathways related to cytokine–cytokine receptor interaction. Overall, modifiable risk factors differ in their relative contributions to cardiovascular disease burden among adults with hypertension. Beyond blood pressure, nicotine exposure showed relatively larger attributable fractions across several outcomes. Multi‐omics analyses point to inflammation‐related pathways as potential biological correlates.
ABSTRACT With the rapid advancement of genome‐wide association studies (GWAS), downstream analyses of GWAS data have become essential for elucidating the genetic mechanisms that underlie complex diseases. However, current post‐GWAS analyses face numerous challenges, including heterogeneous data formats, challenges in multi‐omics integration, and increasingly complex analytical workflows. As a comprehensive post‐GWAS analysis software platform, Omics GWAS provides researchers with an automated, end‐to‐end solution spanning data preparation through results’ visualization by integrating functional modules such as data standardization, Mendelian randomization, multi‐omics joint analysis, comorbidity mechanism exploration, and drug target discovery. The platform supports conversion of GWAS summary statistics across diverse data sources, integrates multidimensional omics data including expression quantitative trait loci (eQTLs), protein quantitative trait loci (pQTLs), and methylation quantitative trait loci (mQTLs) and systematically investigates causal relationships between genotypes and phenotypes using methods such as Mendelian randomization, colocalization analysis, and summary data–based Mendelian randomization (SMR). The modular design of Omics GWAS not only significantly improves analytical efficiency and the reproducibility of results but also offers robust technical support for precision medicine research and drug target development.
ABSTRACT Cervical cancer remains a leading cause of cancer‐related morbidity and mortality in Africa, where progress toward WHO's 90–70–90 elimination targets has been limited by structural health inequalities and gaps in prevention services. We conducted a comprehensive analysis of cervical cancer burden across the African Union from 1990 to 2023, using data from the Global Burden of Disease Study 2023. Four key indicators—incidence, prevalence, mortality, and disability‐adjusted life years (DALYs)—were extracted and stratified by region and age group. Estimated Annual Percentage Change (EAPC) was calculated to assess long‐term trends, whereas age‐specific and regional disparities were evaluated. Future burden was projected to 2038 using the Autoregressive Integrated Moving Average (ARIMA) model. Although global cervical cancer rates have declined modestly since 1990, Africa has experienced rising trends across all indicators. From 1990 to 2023, the age‐standardized incidence rate in the African Union rose from 17.41 to 23.53 per 100,000 population (EAPC = 0.67, 95% CI: 0.48–0.87), and DALYs increased from 368.58 to 430.7 per 100,000 (EAPC = 0.24, 95% CI: 0.07–0.41). Central, Southern, and Eastern Africa showed the highest burden, with Southern Africa recording the most rapid increase in incidence (EAPC = 1.54). Age‐specific analysis revealed a concentrated burden among women aged 40–54, with absolute case numbers nearly doubling in this group since 1990. Projections suggest a substantial increase in incidence, prevalence, deaths and DALYs by 2038; and the gap between Africa and the global average is expected to widen further, with incidence rates remaining significantly above the WHO elimination threshold of < 4/100,000. Africa's cervical cancer burden is increasing and diverging from global progress. Structural health system limitations, insufficient HPV vaccination coverage, and low screening uptake are key contributors. Urgent, region‐specific interventions—integrating prevention, screening, and treatment—are essential to reverse current trends and achieve WHO elimination goals.
ABSTRACT Osteoarthritis (OA) is a common joint disease characterized by degenerative lesions of articular cartilage, subchondral osteosclerosis, osteophyte formation, and synovial inflammation. The traditional view is that OA is a degenerative disease, mainly related to mechanical damage to articular cartilage caused by age, obesity, trauma and other factors. However, in recent years, increasingly more studies have shown that immunoinflammatory responses play a crucial role in the occurrence and development of OA. There are a variety of immune cell infiltration in the joint cavity of OA, and numerous inflammatory factors are released, forming a chronic low‐grade inflammatory microenvironment, promoting articular cartilage degradation and synovial inflammation, and ultimately leading to the occurrence and development of OA. This article reviews the latest research progress on immuno‐inflammatory mechanisms such as immune cell infiltration and inflammatory factor release in OA, and discusses potential strategies for targeted immune response therapy for OA, to provide new ideas for the prevention and treatment of OA.
ABSTRACT With multi‐omics approaches increasingly employed to investigate multidimensional biological systems, current workflows often lack sufficient integration with the underlying genome, limiting comprehensive understanding of genomic structure and chromosomal preferences of candidate genes. Here, we present GAnnoViz (https://github.com/benben‐miao/GAnnoViz/), a cloud‐integrated framework for chromosome‐level feature annotation and visualization in biomedical multi‐omics, encompassing 30+ analytical and visualization modules across five key areas: acquisition of species genomic annotations, extraction of genomic features, drawing of gene structures, sliding window annotation, and chromosome‐scale visualization of multi‐omics. It supports genomic annotation retrieval for 349 species from Ensembl, GFF3 or GTF format compatibility, and an interactive interface deployed on the HiPlot cloud platform. Through three representative application scenarios, we demonstrated GAnnoViz's flexibility in gene structure exploration and utility in annotating and visualizing transcriptomic and epigenomic results, bridging multi‐omics findings with genome architecture for comprehensive biological interpretation at the chromosome scale.
ABSTRACT The escalating global burden of diabetes has led to a rise in diabetic peripheral neuropathy (DPN) that is characterized by significant heterogeneity, with its various subtypes exhibiting distinct yet overlapping clinical manifestations and pathogenic drivers. This complexity obscures a unified mechanistic understanding and hinders the development of precise subtype‐specific therapeutic strategies. This review reframed DPN into a unified model centered on three progressive pathogenic stages: metabolic dysregulation, chronic inflammation, and overt neuronal damage. Within this framework, six DPN subtypes—hyperglycemia‐driven, dyslipidemia‐driven, inflammation, dysvascularity‐driven, small‐fiber DPN, and large‐fiber DPN—were classified into these identified mechanistic triads that converge to redox imbalance. Risk factors across genetic, epigenetic, and phenotypic levels were identified, emphasizing metabolic drivers as key initiators of neuroinflammatory and neurodegenerative cascades. Current therapeutic strategies were evaluated according to their alignment with specific pathogenic stages, ranging from metabolic regulation and anti‐inflammatory approaches to nerve repair. Furthermore, cold atmospheric plasma (CAP) was introduced as a novel multimodal intervention capable of simultaneously targeting all three pathological phases through redox modulation, offering a promising avenue for precision medicine in DPN management. This integrated staging model not only clarified the mechanistic continuum of DPN but also facilitated the development of targeted stage‐specific therapeutic strategies.
ABSTRACT Early relapse represents the most adverse prognostic event in pediatric T cell acute lymphoblastic leukemia (T‐ALL). Clinically, early relapse of pediatric T‐ALL is generally defined as relapse occurring within the first 18 months post‐diagnosis, yet its temporal definition has largely relied on clinical convention rather than quantitative evidence. We systematically analyzed relapse dynamics in three independent cohorts: the CHCMU, AALL03B1, and AALL08B1. Hazard functions, cumulative incidence, and time‐dependent hazard ratios were estimated to characterize relapse risk over time. Landmark Kaplan–Meier analyses and univariate logistic regression were performed to evaluate prognostic impact and molecular determinants of relapse. Across all cohorts, the hazard of relapse peaked around the first year after treatment initiation and declined thereafter, whereas cumulative incidence curves plateaued early. Time‐dependent hazard ratios diverged sharply around 10 months and stabilized around 15 months, identifying a high‐risk subgroup. Landmark analyses at 15 months consistently showed that patients relapsing within this window experienced significantly inferior survival compared with those relapsing later or remaining in remission. Univariate logistic regression analysis identified early relapse as significantly associated with recurrent genetic alterations in NOTCH1, PTEN, IL7R, MYC enhancers, and epigenetic regulators. By integrating temporal risk metrics and external validation, this study provides evidence for POD ≤ 15 months as a data‐driven definition of early relapse in pediatric T‐ALL. This threshold captures patients with markedly poor outcomes, is biologically coherent, and provides a clinically actionable framework for risk‐adapted surveillance, therapeutic stratification, and trial design.
ABSTRACT Linking obesity to mortality is an intriguing and controversial topic. This study tried to comprehensively assess 8 adiposity surrogates and mortality association among middle‐to‐old‐aged adults to identify a superior one, and explore explanatory disorders. Data sources included the National Health and Nutrition Examination Survey (NHANES), UK Biobank (UKB), and FinnGen R11. Cox proportional hazard model, restricted cubic spline, and Mendelian randomization (MR) were used to investigate adiposity surrogates and mortality associations. Observational and MR‐based phenome‐wide association studies (PheWASs) were performed to investigate explanatory disorders. Progressions of disease free‐disease‐death transition were assessed using multi‐state Markov (MSM) models. The NHANES and UKB analyses involved 11,706 and 352,980 participants, respectively. A body shape index (ABSI) showed consistent, stronger, and monotonically‐increasing associations with mortalities in both cohorts. MR analyses demonstrated a monotonically‐increasing causal effect of ABSI on all‐cause mortality (odds ratios [OR] = 1.370, 95% confidence interval [CI]: 1.119–1.676; Pnonlinear = 0.269), which was greater than body mass index (OR = 1.279, 95% CI: 1.163–1.408; Pnonlinear = 0.433). Two PheWASs identified 668 ABSI‐associated disorders, notably atherosclerosis, non‐alcoholic fatty liver disease (NAFLD), type 2 diabetes (T2D), and hypercholesterolemia. Markov models revealed ABSI's role in driving transitions from disease‐free to morbidity of the 4 diseases and death. ABSI, a visceral fat indicator, showed comparative competence in mortality evaluation. The adiposity‐mortality etiology involved diverse diseases across multiple systems, notably atherosclerosis, NAFLD, T2D, and hypercholesterolemia. These findings provide more evidence facilitating development and implementation of clinical guidelines for obesity assessment.
ABSTRACT Sepsis remains a leading cause of mortality worldwide due to its profound biological heterogeneity. While multi‐omics approaches offer promise for precision medicine, existing datasets often lack longitudinal granularity or comprehensive clinical integration. We present the Chinese Multi‐omics Advances In Sepsis (CMAISE) cohort, a prospective, multicenter initiative integrating high‐resolution multi‐omics (transcriptomics, proteomics, metabolomics, and scRNA‐Req) with deep clinical phenotyping. By combining molecular trajectories with continuous vital signs and laboratory data, CMAISE provides an unprecedented resource to decode sepsis complexity. This correspondence outlines the cohort design, data resources, and its potential to accelerate the discovery of treatable traits.
ABSTRACT White adipose tissue browning, by activating core molecules such as PRDM16 and UCP1 to regulate energy metabolism, exhibits a double‐edged sword effect in metabolic diseases and tumors, making it a current research hotspot. This paper systematically reviews its biological characteristics and regulatory networks, including multi‐pathway mechanisms mediated by hormones, transcription factors, noncoding RNAs, and metabolites. It summarizes endogenous and exogenous induction methods such as cold exposure, local hyperthermia, electroacupuncture, exercise, diet, and gut microbiota. Research indicates that GLP‐1 can serve as a therapeutic target for obesity and metabolic syndromes by enhancing thermogenesis, improving insulin resistance, and modulating inflammation. However, within the tumor microenvironment, it may either inhibit tumors through nutrient competition or accelerate cachexia progression, with specific effects varying by tumor types and stages. This paper further explores clinical translation strategies such as GLP‐1 receptor agonists and photothermal nanomaterials, analyzing technical bottlenecks, species differences, and ethical challenges. It emphasizes the need for precise timing of interventions based on the body's energy state. Future research should clarify molecular mechanisms, optimize targeting strategies, and advance its safe translation into novel therapeutic approaches for metabolic diseases and tumors.
