1. The biological age model for evaluating the degree of aging in centenarians.
- Author
-
Zhang, Weiguang, Li, Zhe, Niu, Yue, Zhe, Feng, Liu, Weicen, Fu, Shihui, Wang, Bin, Jin, Xinye, Zhang, Jie, Sun, Ding, Li, Hao, Luo, Qing, Zhao, Yali, Chen, Xiangmei, and Chen, Yizhi
- Subjects
- *
GLOMERULAR filtration rate , *SUPPORT vector machines , *CENTENARIANS , *MULTIPLE regression analysis , *MACHINE learning , *RANDOM forest algorithms , *BODY surface area , *AGING , *THEORY , *FACTOR analysis , *STATISTICAL models , *NUTRITIONAL status - Abstract
• This is the first study of biological age in large-scale centenarians, it success bulid the biological age model in centenarian. • BA models were constructed using traditional methods include multiple linear regression (MLR), principal component analysis (PCA), Klemera and Doubal method (KDM), and machine learning method include srandom forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and light gradient boosting machine (lightGBM) methods. • The aging markers for construction and application including estimated glomerular filtration rate, albumin, pulse pressure, calf circumference, body surface area, fructosamine, and complement 4 cover multiple important organs and systems. Biological age (BA) has been used to assess individuals' aging conditions. However, few studies have evaluated BA models' applicability in centenarians. Important organ function examinations were performed in 1798 cases of the longevity population (80∼115 years old) in Hainan, China. Eighty indicators were selected that responded to nutritional status, cardiovascular function, liver and kidney function, bone metabolic function, endocrine system, hematological system, and immune system. BA models were constructed using multiple linear regression (MLR), principal component analysis (PCA), Klemera and Doubal method (KDM), random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and light gradient boosting machine (lightGBM) methods. A tenfold crossover validated the efficacy of models. A total of 1398 participants were enrolled, of whom centenarians accounted for 49.21%. Seven aging markers were obtained, including estimated glomerular filtration rate, albumin, pulse pressure, calf circumference, body surface area, fructosamine, and complement 4. Eight BA models were successfully constructed, namely MLR, PCA, KDM1, KDM2, RF, SVM, XGBoost and lightGBM, which had the worst R2 of 0.45 and the best R2 of 0.92. The best R2 for cross-validation was KDM2 (0.89), followed by PCA (0.62). In this study, we successfully applied eight methods, including traditional methods and machine learning, to construct models of biological age, and the performance varied among the models. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF