Biomarkers help us understand how cellular and systemic aging contribute to mortality: a study utilizing a machine-learning approach in the Health and Retirement Study.

XGBoost machine learning was used to create system-specific mortality risk scores from biomarkers representing multiple biological systems in a nationally representative sample of older US adults.

• Data came from the Health and Retirement Study (HRS), a nationally representative sample of approximately 4000 US adults over age 55.
• The model was trained in a training subsample using eXtreme Gradient Boosting (xgboost).
• Biological systems modeled included the brain and nervous system, adaptive immune system, cardiovascular system, and renal system, as well as general multisystem aging.
• The approach produced system-specific mortality risk scores based on sets of biomarkers representing each biological system.

The effects of most biological systems on mortality may be well captured by one or a small number of biomarkers.

• This finding emerged from the XGBoost feature importance analysis across system-specific models.
• Rather than requiring comprehensive multi-biomarker panels, individual systems tended to have one dominant biomarker driving mortality prediction.
• This suggests parsimony in biomarker selection may be adequate for representing specific physiological systems in aging research.

Female sex was found to be either a protective or a risk factor for mortality depending on the specific biological system examined.

• The direction of the association between female sex and mortality risk varied across different biological system-specific models.
• This finding indicates that sex differences in aging are system-specific rather than uniformly protective or detrimental.
• The result highlights the importance of examining sex differences within specific physiological systems rather than assuming a global protective effect of female sex.

General biological aging and system-specific aging were both identified as important and potentially distinct contributors to health outcomes.

• The study examined both general multisystem aging and system-specific aging as separate constructs.
• How general biological aging and specific systemic aging co-occur and influence one another to affect health outcomes was described as 'largely unknown' prior to this work.
• The authors discuss the importance of studying both general and system-specific aging simultaneously.

The study provides a framework for understanding how social exposures may differentially accelerate decline in individual physiological systems.

• The paper identified emerging interest in understanding how social exposures may differentially accelerate decline in individual physiological systems.
• System-specific mortality risk scores allow for examination of which biological systems may be more susceptible to social and environmental influences.
• This approach enables decomposition of overall biological aging into system-level components that may respond differently to social determinants.