Background: The process of aging and metabolism are intricately linked, thus rendering the identification of reliable biomarkers related to metabolism crucial for delaying the aging process. However, research of reliable markers that reflect aging profiles based on machine learning is scarce.
Methods: Serum samples were obtained from aged mice (18-month-old) and young mice (3-month-old). LC-MS was used to perform a comprehensive analysis of the serum metabolome and machine learning was used to screen potential aging-related biomarkers.
Results: In total, aging mice were characterized by 54 different metabolites when compared to control mice with criteria: VIP ≥ 1, q-value < 0.05, and Fold-Change ≥ 1.2 or ≤0.83. These metabolites were mostly involved in fatty acid biosynthesis, cysteine and methionine metabolism, D-glutamine and D-glutamate metabolism, and the citrate cycle (TCA cycle). We merged the comprehensive analysis and four algorithms (LR, GNB, SVM, and RF) to screen aging-related biomarkers, leading to the recognition of oleic acid. In addition, five metabolites were identified as novel aging-related indicators, including oleic acid, citric acid, D-glutamine, trypophol, and L-methionine.
Conclusions: Changes in the metabolism of fatty acids and conjugates, organic acids, and amino acids were identified as metabolic dysregulation related to aging. This study revealed the metabolic profile of aging and provided insights into novel potential therapeutic targets for delaying the effects of aging.
Keywords: aging; biomarkers; machine learning; metabolomics.