Scholarlink

Background <p>Osteoporosis (OP) is a prevalent metabolic bone disorder and a major public health concern characterized by reduced bone mass and bone microstructural deterioration. Early identification of osteoporosis and implementation of preventive interventions remain critical for reducing fracture risk and disease burden. Identifying plasma biomarkers reflecting metabolic alterations related to OP may facilitate early detection and risk assessment.</p> Methods <p>&#xa0;Untargeted metabolomics and lipidomics profiling based on ultra-performance liquid chromatography-mass spectrometry (UHPLC-MS) was performed in a discovery cohort comprising 75 patients with OP and 140 healthy controls. Differential features were identified using multivariate statistical analysis (PCA, OPLS-DA), and FDR-adjusted univariate analysis. Multivariable logistic regression and LASSO-regularized logistic regression was employed, with age and sex incorporated as mandatory covariates to identify independent lipid predictors. A Random Forest (RF) model was further evaluated in an independent validation cohort consisting of 20 OP patients and 42 healthy controls.</p> Result <p>&#xa0;Sixty-one differential metabolites were identified, primarily enriched in lipid metabolism pathways. Further targeted lipidomics identified four diagnostic lipid biomarkers, including LPA(16:0), LPI(16:0), LPI(18:0), and LPI(20:0). Following covariate adjustment for age and sex, key lipid species remained independently associated with OP. The RF-based diagnostic model maintained robust performance in the validation cohort, yielding an AUC of 0.916 (95% CI: 0.842–0.990), with high sensitivity and specificity.</p> Conclusions <p>LPA(16:0), LPI(16:0), LPI(18:0), and LPI(20:0) in plasma were negatively correlated with the T value of bone mineral density. These associations persisted after stringent adjustment for age and sex, suggesting that lysophospholipid dysregulation is an independent metabolic hallmark of OP. The multi-metabolites model based on four biomarkers showed promising predictive performance for OP and may provide a potential tool for early risk assessment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integration of metabolomics and machine learning algorithm for discovery of early diagnostic biomarkers of osteoporosis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月14日

查看全文
  • Jing Liu,
  • Jialong Wang,
  • Zhijun Bao,
  • Jingkun Jia,
  • Jinru Jia,
  • Lifeng Han,
  • Jinghui Du,
  • Erwei Liu

摘要

Background

Osteoporosis (OP) is a prevalent metabolic bone disorder and a major public health concern characterized by reduced bone mass and bone microstructural deterioration. Early identification of osteoporosis and implementation of preventive interventions remain critical for reducing fracture risk and disease burden. Identifying plasma biomarkers reflecting metabolic alterations related to OP may facilitate early detection and risk assessment.

Methods

 Untargeted metabolomics and lipidomics profiling based on ultra-performance liquid chromatography-mass spectrometry (UHPLC-MS) was performed in a discovery cohort comprising 75 patients with OP and 140 healthy controls. Differential features were identified using multivariate statistical analysis (PCA, OPLS-DA), and FDR-adjusted univariate analysis. Multivariable logistic regression and LASSO-regularized logistic regression was employed, with age and sex incorporated as mandatory covariates to identify independent lipid predictors. A Random Forest (RF) model was further evaluated in an independent validation cohort consisting of 20 OP patients and 42 healthy controls.

Result

 Sixty-one differential metabolites were identified, primarily enriched in lipid metabolism pathways. Further targeted lipidomics identified four diagnostic lipid biomarkers, including LPA(16:0), LPI(16:0), LPI(18:0), and LPI(20:0). Following covariate adjustment for age and sex, key lipid species remained independently associated with OP. The RF-based diagnostic model maintained robust performance in the validation cohort, yielding an AUC of 0.916 (95% CI: 0.842–0.990), with high sensitivity and specificity.

Conclusions

LPA(16:0), LPI(16:0), LPI(18:0), and LPI(20:0) in plasma were negatively correlated with the T value of bone mineral density. These associations persisted after stringent adjustment for age and sex, suggesting that lysophospholipid dysregulation is an independent metabolic hallmark of OP. The multi-metabolites model based on four biomarkers showed promising predictive performance for OP and may provide a potential tool for early risk assessment.