Paulo Jorge Gavaia

A laboratory- and imaging-free machine-learning model for osteoporosis, trained on U.S. adults, was applied frozen to a Korean cohort. Discrimination transferred almost unchanged (AUC 0.79), needing only a light recalibration, showing that a DXA-independent screening tool can generalise across populations differing markedly in ancestry, body habitus and survey instrumentation.

Purpose: The application of machine learning (ML) to osteoporosis prediction has expanded rapidly, yet no comprehensive meta-analysis has synthesized the discriminative performance of these models across all ML categories, data types, and validation strategies. This systematic review and meta-analysis aimed to evaluate the diagnostic and predictive accuracy of ML and deep learning models for osteoporosis prediction in adult populations. Methods: Systematic searches of PubMed, Embase, Web of Science, and IEEE Xplore were conducted for studies published between January 2020 and February 2026. Studies developing, validating, or applying ML models for predicting osteoporosis, low bone mineral density, or osteoporotic fractures in adults were included. Methodological quality was assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST). Area under the receiver operating characteristic curve (AUC) values were pooled using random-effects meta-analysis with logit transformation. Subgroup analyses were performed by data type, ML category, external validation status, and population type. The review followed PRISMA 2020 guidelines. Results: Thirty-three studies were included in the qualitative synthesis and 27 in the meta-analysis. The pooled AUC was 0.879 (95% CI: 0.853–0.901), with substantial heterogeneity (I² = 99.5%). Imaging-based models outperformed clinical data models (AUC = 0.905 vs. 0.872). Deep learning achieved the highest pooled AUC (0.909), followed by ensemble methods (0.874) and traditional ML (0.840). Externally validated models showed lower performance than internally validated ones (AUC = 0.868 vs. 0.897). PROBAST assessment rated 32 of 33 studies (97.0%) as low risk of bias, though this proportion should be interpreted cautiously given that PROBAST was designed for traditional prediction models and may not fully capture ML-specific sources of bias. Egger’s test indicated significant publication bias (p < 0.001). Explainable AI methods were employed in 60.6% of studies, identifying age, body weight, and alkaline phosphatase as the most frequent top predictive features. Conclusions: Machine learning models demonstrate overall good discriminative performance for osteoporosis prediction, albeit with substantial heterogeneity across studies (I² = 99.5%), and show potential as complementary screening tools, particularly in settings with limited DXA access. Deep learning models applied to imaging data and ensemble methods using clinical variables achieved the strongest subgroup estimates. However, extreme heterogeneity, evidence of publication bias, and limited prospective validation warrant cautious interpretation of the pooled estimate. Future research should prioritise multi-centre external validation, standardised reporting following TRIPOD+AI guidelines, and prospective clinical trials to establish real-world clinical impact. Mini- This meta-analysis of 33 studies demonstrates that ML models achieve a pooled AUC of 0.879 for osteoporosis prediction, with imaging-based deep learning models reaching 0.905, supporting their potential as complementary screening tools to DXA and FRAX®.

Purpose/IntroductionOsteoporosis is characterized by diminished BMD and deteriorated bone microstructure, significantly increasing fracture susceptibility. This study leverages machine learning for identifying key predictors of osteoporosis, potentially improving non-invasive screening tools.MethodsWe utilized the "Bone Mineral Density" dataset from Harvard University, categorizing variables into demographics, bone loss prevention treatments, and lifestyle/associated pathologies. A Boolean variable "OP" was defined to indicate presence of osteoporosis. Data preprocessing included imputation for missing values and normalization. Statistical analyses such as Welch's T-test and multiple linear regression were conducted to identify significant markers. Additionally, unsupervised learning techniques, like Louvain clustering, were used to uncover natural data patterns, while supervised learning models aimed to predict osteoporosis.ResultsThe study identified significant markers associated with osteoporosis, including magnesium levels, high- and low-density lipoproteins, chronic obstructive pulmonary disease, and chronic kidney disease. Multiple linear regression analysis revealed robust associations between osteoporosis and biomarkers such as ALT, BUN, CREA, LDL-C, and Mg, even after controlling for age, height, and weight. Among supervised learning models, Random Forest and Gradient Boosting algorithms showed the highest accuracy in predicting osteoporosis, with Random Forest identified as the preferred model.ConclusionsIn our work, it was possible to identify the main biochemical parameters and clinical histories that are potential predictors of osteoporosis. The findings suggest that integrating machine learning techniques with clinical data can enhance early detection and intervention strategies, ultimately improving patient outcomes. Further refinement is recommended to validate these results across diverse populations and improve predictive models for broader application.

Endoperoxide-containing molecules based on the antimalarial drug artemisinin have demonstrated various biological properties, including modulation of calcium homeostasis. This study evaluated the toxicity and osteogenic activity of five newly developed tetraoxanes (YB1, YB9, YB11, YB17 and T2), alongside three of their non-peroxidic analogues (IC22, IC26 and IC33), in zebrafish (Danio rerio) larvae. For each compound, LC₅₀ values were first determined. Behavioural responses and morphological alterations were studied as indicators of toxicological impact. The osteogenic activity was assessed through the operculum assay, followed by the analysis of gene expression markers related to calcium homeostasis (atp2a1), oxidative stress (sod1, cat), and osteogenesis (sp7, oc2). All the compounds evaluated induced an inhibition of osteogenic activity. T2, YB11, IC33 and IC26 affected the locomotor function by decreasing swimming activity. IC26 and IC33 induced morphological toxicity, characterized by a curved trunk and alterations in larval body curvature. From all the compounds studied, YB1, YB9, YB17 and IC22 showed selective anti-osteogenic activity, without displaying significant behavioural or morphological toxicity. In conclusion, the presence of a peroxide bond in the molecular structure of the compounds increases the anti-osteogenic activity at lower concentrations. All evaluated compounds exhibited anti-osteogenic activity and can be regarded as anti-osteogenic agents. However, YB17 did not induce transcription alterations in the genes analyzed and may thus represent the most promising compound in conditions where a controlled inhibition of bone formation is desirable.

Over the past few decades, research in aging-related bone diseases such as osteoporosis and osteoarthritis has made significant strides. Yet, the tangible impact of these scientific advances on the daily lives of both humans and companion animals remains limited. Progress in the area of nutrition have often been hampered and confined within disciplinary borders, resulting in a fragmented body of knowledge. This Research Topic seeks to overcome that fragmentation by adopting a multidisciplinary perspective. Aging bone health is not merely a clinical issue, nor solely a question of molecular biology or nutrient metabolism. It is a complex confluence of factors -genetics, gender, physiological condition, co-morbidities, physical activity, environment, and habitual behaviors -that affect outcomes in a highly individualized manner.The articles in this Research Topic collectively demonstrate the value of integrated research and point to promising paths forward. Several contributions use radiomics and imaging-based predictive models to improve early diagnosis and risk stratification for osteoporosis and related fractures -tools that can be pivotal in lowresource settings such as primary care hospitals. The contribution by Chen et al. (2025) introduce a novel screening tool for osteoporosis. By using a model based on clinical and radiomic features, this toll has the potential for democratizing osteoporosis screening in underserved regions. Yu et al. (2025) present a new lumbar CT-based radiomics model for predicting osteoporotic vertebral compression fractures (OVCF) in postmenopausal women, offering a possible alternative to traditional bone mineral density (BMD) assessments. Notably, this model outperformed DEXA scans in sensitivity and accuracy, even among patients with similar T-scores, highlighting the inadequacy of relying on single metrics in isolation and reinforcing the need for individualized risk assessment.Beyond imaging, biochemical and nutritional biomarkers should also be further explored and integrated into the risk assessment and diagnosis processes. The study by Liu et al. (2025) shows as albumin-corrected anion gap (ACAG) levels correlate with reduced lumbar spine BMD, suggesting a novel metabolic marker for osteoporosis risk -though the relationship is nonlinear.The study by Li et al. (2023) further deepens the nutritional lens on aging and bone health: the BMD of the lumbar spine was positively correlated with serum 25hydroxyvitamin D levels, the difference between actual body weight and ideal body weight, and uric acid levels in the blood. Conversely, the BMD of the femoral neck was negatively correlated with age and female sex, and positively correlated with height and geriatric nutrition risk index score, suggesting a protective role of uric acid over bone in advanced age, and underscoring the necessity for personalized dietary recommendations.The article by Guo et al. ( 2024) found sex-and bone-site-specific variations in the associations between vitamin K intake levels and bone health in individuals aged over 50 years.The intersection of bone health with cardiovascular and muscular health adds further layers to the picture.The study by Westbury et al. (2025) illustrates how calcium supplementationcommonly recommended for bone health -may paradoxically increase fracture risk, possibly due to reverse causality, while dietary calcium shows protective effects against cardiovascular mortality. Whilst dietary calcium intake was not associated with either overall fracture or hip fracture, a better-quality diet was related to reduced risk of hip fracture, after adjustment for sex. Such findings call for more nuanced dietary guidelines that go beyond isolated nutrients and consider longterm patterns and holistic outcomes. Importantly, prudent diet scores were also associated with beneficial lifestyle traits, such as higher physical activity and lower rates of smoking, reflecting the broader complexity of the socio-behavioral ecosystem in which nutrition operates. The systematic review and metanalysis by Yan et al. (2025) identified osteoporosis as a significant risk factor for development of stroke related sarcopenia, along with age, tube feeding, pre-stroke sarcopenia, atrial fibrillation, neurological compromise ( as evaluated by the NIHSS score).Finally, the study by Liu et al (2024) explores the relationship between the use ofdiuretics and fracture risk. Diuretics are used widely in the treatment of conditions such as hypertension, cardiac and kidney disease. This study specifically approaches the differing effects of loop, thiazide, and potassium-sparing diuretics, reinforcing the critical need for integrated pharmacological and nutritional strategies, and further research, especially in aging populations vulnerable to falls and bone injuries.Taken together, the articles in this Research Topic demonstrate that advancing bone health in aging requires a system-level approach, grounded in real-world variability.We encourage clinicians, researchers, veterinarians, nutritionists, and public health professionals to embrace a framework that bridges disciplines and actively translates findings into interventions tailored to individual contexts.The editors would like to thank the excellent contributions of the reviewers and editorial team to the topic here presented.We hope that this collection will provide meaningful insights to the development of practical, evidence-based tools and strategies that are responsive to complexity, rather than reductive. Only through such an approach can we move from fragmented insights to holistic solutions-benefiting the health and longevity of the living beings we care for.

Background: Targeting immune pathways to prevent bone loss represents a promising, yet underexplored therapeutic strategy.Methods: An ethanol-soluble fraction derived from the freeze-dried biomass of the marine microalga Skeletonema costatum (SKLT) was tested for its ability to modulate immune responses and inhibit osteoclastogenesis. Its effects were evaluated in a zebrafish model of bone regeneration, a medaka model of RANKL-induced osteoporosis, and in vitro using murine RAW 264.7 macrophages. Transcriptomic profiling of regenerating fin blastemas at 24 hours post-amputation was performed to identify the affected molecular pathways.Results: In zebrafish, SKLT treatment suppressed the recruitment of osteoclast precursors and altered mineralization dynamics. Transcriptomic profiling revealed downregulation of genes involved in inflammation, antigen presentation, T-cell activation, and macrophage commitment towards osteoclastogenesis, accompanied by reduced expression of chemokines and cytokines that promote osteoclast precursor recruitment and fusion. In medaka, SKLT significantly reduced vertebral bone loss and enhanced neural arch mineralization in larvae with high RANKL expression. In vitro, SKLT inhibited proliferation and osteoclastic differentiation of murine RAW 264.7 macrophages exposed to RANKL without inducing cytotoxicity.Conclusion: These findings identify S. costatum as a source of bioactive immunomodulatory compounds capable of interfering with key osteoimmune mechanisms. Beyond providing proof of concept for their therapeutic potential in bone erosive disorders, this work opens avenues for isolating and characterizing the active molecules, optimizing their delivery, and evaluating their efficacy in preclinical mammalian models. Such strategies could expand the repertoire of safe, nutraceutical-based or adjuvant therapies for osteoporosis and other inflammation-driven skeletal diseases, complementing and potentially enhancing current antiresorptive and anabolic treatments.

Flatfish (Pleuronectiformes) encompass fascinating fish species known for their high gastronomical value, rearing performances, and market prices, which attract both economic and scientific interest. Despite these attributes, flatfish exhibit unique morphological, developmental, and biological features, as well as distinctive swimming movements, which require special approaches in skeletal biology research, an area that remains relatively unexplored for this diverse group of fishes. These unique features emerge during metamorphosis, at the end of the larval period, when bilateral symmetric pelagic larvae gradually become asymmetric and undergo various morphological and physiological changes. Consequently, the skeletal characteristics of these species include a certain degree of asymmetry in the skeletal structures. These fish also present acellular bone (bone devoid of osteocytes). Research on flatfish skeletons is crucial for improving animal welfare and promoting sustainable farming, as the literature indicates that more than 50% of the larvae and juveniles of different cultured flatfish can be affected by skeletal deformities. This review aims to compile the available studies on the development of vertebral column deformities in flatfish in an integrative manner, addressing state‐of‐the‐art research on the etiology, diagnostics, and innovative studies on this issue. It also covers the fundamental aspects of teleost skeletal development, particularly the flatfish vertebral column.

Epidemiologic studies reveal that the fight against diabetes is being lost, so to overcome this serious health problem every strategy should be taken seriously. Using in silico data we discuss the therapeutic potential for Diabetes Mellitus type 1 of two possible unstudied strategies; (I) if INSULIN paralog, INSULIN-INSULIN GROWTH FACTOR 2 can be a regulator of glucose in humans, and (II) if extrapancreatic expression of INSULIN and INSULIN- INSULIN GROWTH FACTOR 2 has any relevance in glucose regulation. To address this hypothesis, we access to the collection of protein sequences and gene expression from INSULIN and paralogues in EGB, NCBI and EBI. Phylogenetic and molecular evolutionary analyses were conducted using MEGA software, conserved protein motifs were compared using algorithm MUSCLE and presenting it graphically using CLC Genomics Workbench, transcripts and gene expression from transcriptome and microarray data from different experiments of insulin and paralogues were extracted and graphicaly presented using CIRCOS. Phylogenetic analysis revealed that humans, like most of all mammals, have INSULIN and INSULIN-INSULIN GROWTH FACTOR 2 with the exception of murinae among mammals that share Insulin 1 and Insulin 2 . Both Insulin 1 and 2 can regulate glucose since both proteins are highly similar and with a redundant function, as confirmed by in vivo experiments showing that both knockout Insulin 1 or 2 are normoglycemic. Human INSULIN- INSULIN GROWTH FACTOR 2 lacks motif 3 of the INSULIN signature, and part of the predicted protein structure does not present any known domain, making the prediction of a possible function for this gene inconclusive. Extrapancreatic expression of insulin has been observed in diabetic mice with contradictory effects, but available gene expression data in repository data banks also suggests that extrapancreatic expression of insulin occurs in human, mouse and zebrafish under basal condition. Gene expression DataSets demonstrate that extrapancreatic expression of INSULIN occurs under basal condition. Further investigation should be made, to understand if extrapancreatic expression of insulin can have a role in glucose homeostasis, as this could represent a complementary strategy in diabetes treatment.

Purpose/Introduction: Osteoporosis is characterized by diminished BMD and deteriorated bone microstructure, significantly increasing fracture susceptibility. This study leverages machine learning for identifying key predictors of osteoporosis, potentially improving non-invasive screening tools. Methods: We utilized the “Bone Mineral Density” dataset from Harvard University, categorizing variables into demographics, bone loss prevention treatments, and lifestyle/associated pathologies. A Boolean variable “OP” was defined to indicate presence of osteoporosis. Data preprocessing included imputation for missing values and normalization. Statistical analyses such as Welch’s T-test and multiple linear regression were conducted to identify significant markers. Additionally, unsupervised learning techniques, like Louvain clustering, were used to uncover natural data patterns, while supervised learning models aimed to predict osteoporosis. Results: The study identified significant markers associated with osteoporosis, including magnesium levels, high- and low-density lipoproteins, chronic obstructive pulmonary disease, and chronic kidney disease. Multiple linear regression analysis revealed robust associations between osteoporosis and biomarkers such as ALT, BUN, CREA, LDL-C, and Mg, even after controlling for age, height, and weight. Among supervised learning models, Random Forest and Gradient Boosting algorithms showed the highest accuracy in predicting osteoporosis, with Random Forest identified as the preferred model. Conclusions: In our work, it was possible to identify the main biochemical parameters and clinical histories that are potential predictors of osteoporosis. The findings suggest that integrating machine learning techniques with clinical data can enhance early detection and intervention strategies, ultimately improving patient outcomes. Further refinement is recommended to validate these results across diverse populations and improve predictive models for broader application.