The natural aging process causes cellular and molecular damage that results in reduced body functions and higher vulnerability to Alzheimer’s disease, Parkinson’s disease, and osteoporosis. The treatment of these diseases remains challenging, yet proper diagnosis at an early stage combined with successful management techniques can help to slow their progression.
The fundamental process of understanding aging depends on calculating biological age (BA) because this measurement diverges from chronological age (CA) to indicate wellness status. Researchers have yet to overcome the challenge of BA prediction, as both genetic and environmental factors influence aging patterns. Studies on biological age assessment have prompted researchers to develop complex models that investigate biological indicators.
A deep learning model that relies on steroid metabolic pathways was developed by researchers to determine biological aging predictions from steroid hormones. The research selected 150 healthy test subjects ranging from 20 to 73 years old to provide blood serum samples that underwent steroid measurement tests. The researchers utilized 30 steroid standards together with 14 internal standards to understand the relationship between steroid synthesis and aging parameters. An analysis using deep neural networks processed these steroid profiles to simulate how steroids affect biological aging. Researchers tested the model on 100 participants first while validating it by testing 50 participants, including 10 smokers, to determine smoking-related BA effects.
The DNN model delivered crucial knowledge about the connection that exists between steroid metabolism and the process of aging. The prediction model accurately determined BA by evaluating the complicated biological systems that lead to aging processes. The study proved that steroid analytics of corticosteroids and sex hormones serve as essential methods to explain aging disparities between individuals. The model data revealed notable differences in BA forecasting capability between male and female participant samples because metabolic processing functions differently for males and females. The aging rate for male smokers increased when their steroid levels escalated, thus demonstrating how life choices influence the acceleration of aging.
The process of steroid metabolism functions throughout adrenal glands, liver, and gonads while promoting aging throughout the entire body system. This model established connections between individual organ transformations and general aging development through steroid profile assessment, which served as a real-time measure for biological age evaluation. When researchers applied cumulative distribution function(CDF)-based proportional scaling to their steroid concentration measurements, they achieved better data consistency, thus reducing errors from both experimental and biological factors.
The research demonstrates why steroid metabolism plays a crucial role in biological aging. Using the DNN model, researchers achieved successful integration of steroid pathways in addition to metabolic factors for predicting BA, which provided a new understanding of how hormones like cortisol (COL) reflect stress and cumulative physiological damage. The research analysis confirmed that COL has a positive relationship with BA, strengthening its value as an aging biomarker. The model discovered that smoking, as well as other lifestyle factors, affected BA rates while identifying smoking as particularly detrimental because male smokers moved toward premature aging faster than non-smokers. Lifestyle changes emerge as vital components for handling the natural process of aging.
However, there are several limitations. The study’s relatively small sample size and lack of detailed lifestyle data in some cohorts may reduce the generalizability of the results. Additionally, the model’s dependence on a specific set of steroids may limit its applicability to datasets that do not include a full steroid panel. The model did not include patterns of circadian rhythms as factors influencing steroid levels across different time periods. Future research needs to evaluate participants throughout time to create dynamic models that effectively monitor steroid profile changes throughout the day.
The study confirms that steroid metabolic pathways hold promise as a method for estimating biological age. Better modeling of age-related processes through this research can facilitate more effective individual treatment strategies for age-related diseases. Further research needs to use more diverse datasets and study environmental and lifestyle factors to enhance the precision of BA prediction models. Advancing such tools will equip medical professionals with valuable methods to assess and improve care for aging patients.
References: Wang Q. Biological age prediction using a DNN model based on pathways of steroidogenesis. Sci Adv. 2025;11:eadt2624. doi:10.1126/sciadv.adt2624
