PREDICTING THE VULNERABILITY AND RESILIENCE TO CARDIOVASCULAR AND NEUROENDOCRINE EFFECTS OF STRESS IN ADULT RATS THROUGH A NOVEL MACHINE LEARNING APPROACH

Abstract

Chronic stress has been risk of cardiovascular disease and neuroendocrine illness in humans and animals. However, not all individuals are equally vulnerable to the negative effects of stress, and some may even exhibit resilience. Identifying biomarkers or other predictors of vulnerability and resilience could help to develop personalized prevention and treatment strategies. In this study, we aimed to predict vulnerability and resilience to stress-related health effects in adult rats using a novel machine learning approach. We exposed male rats to chronic stress or control conditions for six weeks and measured their cardiovascular and neuroendocrine responses at baseline and at the end of the stress exposure. Rats were considered vulnerable if they exhibited large growth in heart rate and reaction of blood pressure to stress, and resilient if they did not show significant changes in these parameters. We then applied a novel machine learning algorithm to identify patterns in the data that could predict vulnerability or resilience. In this case, we employed a combination methods for selecting features using Support Vector Machine and classification algorithm Principal component Analysis to identify the most important predictors of vulnerability and resilience. We also compared the performance of the machine learning approach with traditional statistical methods, such as logistic regression and discriminant analysis. Our results suggest that heart rate variability were among the most important predictor of vulnerability and resilience to stress-related health effects in rats. Specifically, rats with lower heart rate variability and higher cortisol levels at baseline were more likely to be vulnerable to stress. Conversely, rats with greater concentrations of anti-inflammatory cytokines increased risk of becoming resilient to stress. The machine learning approach was more accurate in predicting vulnerability and resilience than traditional statistical methods, with an overall accuracy of 89%, respectively. Our study provides new insights into the complex interplay between stress and health, and highlights the potential of machine learning to improve our understanding of this relationship. The identification of biomarkers and predictors of vulnerability and resilience could lead to the development of personalized approaches to stress management and prevention of stress-related health conditions.