Abstract

Poster Presentations

Day 2, June 11（Tue.）　　Room P1 (Multipurpose Hall)・Room P2 (Conference Room 101+102)

Aging encompasses the progressive accumulation of molecular and cellular damage, gradual organ function decline, and heightened disease susceptibility. Genetic factors contribute to only 20-25% of aging's influence, with environmental and lifestyle factors exerting a greater impact (1). Therefore, it is crucial to devise methods for accurately estimating individuals' biological age, identifying potential biomarkers, and elucidating their specific relationship with aging, vital for disease diagnosis and prognosis (2). While previous research has explored biological age prediction modeling using transcriptomic or metabolomic data, challenges persist in ensuring the biological interpretability of modeling parameters.
In this study, we analyzed 100 serum samples in total from individuals aged 20-70 years, quantifying 22 steroids using an in-house developed, highly sensitive LC-MS/MS detection technique (3). This enabled us to establish a gold standard for normalizing this data category. Employing neural network modeling based on metabolomic pathways from normalized results, we identified potential biomarkers and developed methods for computing biological age. Our findings revealed sex-specific differences in the reciprocal weighting of steroid metabolic pathways in biological age modeling. Additionally, we validated our model by analyzing a smoker cohort, affirming lifestyle habits' impact on biological age using real-world data.

---