Activity dynamics allow early discrimination of infection-related survival outcomes
Journal:
bioRxiv
Published Date:
May 8, 2026
Abstract
Predicting transitions between health, disease, and death across biological systems remains an important challenge with significant implications for both ecological management and medical intervention. Although the principles underlying these transitions are increasingly recognised, accurate and tractable dynamical indicators of health-to-disease transitions remain rare, especially at the level of individual organisms. Here, we use dynamical statistical indicators of high-resolution activity time series to predict infection-related mortality. By analysing locomotor activity data from infected Drosophila melanogaster flies, we find that individual dynamical indicators, such as the mean, variance, autocorrelation, and permutation entropy, differed between flies that survived and those that died during the experiment. When these indicators were used to train a Random Forest model, the classifier performed well (AUC = 0.94), demonstrating an accuracy of 87.9% in discriminating between infected flies that would die from infection and those that would survive, with the strongest discriminatory power detected over 12 hours prior to death. Our findings show that combining these easy-to-compute, dynamical statistical indicators with machine learning enhances the ability to predict health deterioration in the Drosophila model. Conceptually, our findings emphasize that the integration of dynamical statistical metrics from physiological or behavioural time-series with machine learning approaches may offer a promising avenue for real time health monitoring in both ecological and clinical settings.
