# Investigating Stress and Coping Behaviors in African Green Monkeys (Chlorocebus aethiops sabaeus) Through Machine Learning and Multivariate Generalized Linear Mixed Models

## TL;DR

This study uses machine learning and statistical models to explore stress and coping behaviors in African green monkeys, aiming to improve their welfare in captivity.

## Contribution

The novel integration of machine learning and multivariate generalized linear mixed models to analyze stress-related behaviors in African green monkeys.

## Key findings

- Hair cortisol levels did not differ between groups exposed to enrichment activities.
- Statistical analysis revealed significant behavioral patterns linked to physiological stress markers.
- The study highlights the importance of advanced statistical techniques for objective animal welfare assessments.

## Abstract

Understanding both behavior and physical health is important for measuring how well animals are coping in captivity. In this study, we collected hair, blood, and saliva samples from 40 male African green monkeys (AGMs) (Chlorocebus aethiops sabaeus) to measure their stress responses. We used principal component analysis (PCA) with a Bayesian mixed model analysis to find patterns in their behaviors related to cortisol, lysozyme, and β-endorphin. While the animals were divided into two groups to see if an enrichment activity would help their welfare, there was no difference in their hair cortisol levels. The statistical analysis, however, shows certain behaviors were connected to stress, suggesting that we need more research to understand how factors like environmental and social interactions are connected to animal welfare. This study shows that looking closely at animal behaviors with advanced statistical techniques can provide better objective assessments of behavior, which can lead to better veterinary management practices.

Integrating behavioral and physiological assessment is critical to improve our ability to assess animal welfare in biomedical settings. Hair, blood, and saliva samples were collected from 40 recently acquired male African green monkeys (AGMs) to analyze concentrations of hair cortisol, plasma β-endorphin, and lysozyme alongside focal behavioral observations. The statistical methodology utilized machine learning and multivariate generalized linear mixed models to find associations between behaviors and fluctuations of cortisol, lysozyme, and β-endorphin concentrations. The study population was divided into two groups to assess the effectiveness of an enrichment intervention, though the hair cortisol results revealed no difference between the groups. The principal component analysis (PCA) with a Bayesian mixed model analysis reveals several significant patterns in specific behaviors and physiological responses, highlighting the need for further research to deepen our understanding of how behaviors correlate with animal welfare. This study’s methodology demonstrates a more refined approach to interpreting these behaviors that can help improve animal welfare and inform the development of better management practices.