Early detection of stress is crucial to preventing decline in performance, mental and physical health problems, and improving overall well-being. Four stress trials (heat, muscle exertion, sleep deprivation, and psychosocial stress) were conducted to assess the impact of stress on interrelationships between biological, cognitive, and physical states.
This study is unique in its approach, aiming to elucidate the specific relationships between health habits, psychological, cognitive, physiological, and physical aspects and omics measures derived from different biological samples. An integrative analytical pipeline was employed, starting with unsupervised machine learning for variable reduction, followed by Redundancy Analysis (RDA) and supervised machine learning to determine which traits explain the most variance in the full omics datasets. Structural Equation Modelling (SEM) was then utilised to explore causal relationships between these traits and the omics data.
This multi-step approach identified key variables that significantly contributed to the variance in the omics data, revealing distinct patterns and relationships within each biological sample type. The variance across peripheral omics datasets (gut microbiome and miRNA from saliva) was linked to cognitive, physical, and sleep datasets, revealing specific relationships. For example, nap duration, beverage consumption, and viral flu had a greater impact on the gut microbiome than on the miRNA of saliva. This approach offered new perspectives to complement traditional statistical analyses, such as differential expression and has served as a preliminary step before these analyses.
These findings have significant implications, emphasising the importance of integrating diverse data types to comprehensively understand stress and performance. This approach could advance early detection of stress while predicting and improving performance.
Keywords: Anthropologic Health Habits, Microbiome, Injury Prediction, Military, Performance, Random Forest, Principal Component Analysis, Structural Equation Modelling (SEM), Predictive Analytics.