ABSTRACT
BACKGROUND: Menopause, a dramatical estrogen-deficient condition, is considered the most significant milestone in women's health. PURPOSE: To investigate the metabolite changes attributed to estrogen deficiency using random forest (RF)-based machine learning (ML) modeling strategy in ovariectomized (OVX) mice as well as determine the clinical relevance of selected metabolites in older women. METHODS AND RESULTS: Untargeted and targeted metabolomic analyses revealed that metabolites related to TCA cycle, sphingolipids, phospholipids, fatty acids, and amino acids, were significantly changed in the plasma and/or muscle of OVX mice. Subsequent ML classifiers based on RF algorithm selected alpha-ketoglutarate (AKG), arginine, carnosine, ceramide C24, phosphatidylcholine (PC) aa C36:6, and PC ae C42:3 in plasma as well as PC aa 34:1, PC aa C34:3, PC aa C36:5, PC aa C32:1, PC aa C36:2, and sphingosine in muscle as top featured metabolites that differentiate the OVX mice from the sham-operated group. When circulating levels of AKG, arginine, and carnosine, which showed the most significant changes in OVX mice blood, were measured in postmenopausal women, higher plasma AKG levels were associated with lower bone mass, weak grip strength, poor physical performance, and increased frailty risk. CONCLUSIONS: Metabolomics- and ML-based methods identified the key metabolites of blood and muscle that were significantly changed after ovariectomy in mice, and the clinical implication of several metabolites was investigated by looking at their correlation with body composition and frailty-related parameters in postmenopausal women. These findings provide crucial context for understanding the diverse physiological alterations caused by estrogen deficiency in women.