Early warning signals for desynchronization in periodically forced systems

TL;DR

This paper introduces a method to detect early warning signals of desynchronization in biological systems, especially when traditional indicators fail due to the nature of the bifurcation involved.

Contribution

It demonstrates how to identify early warning signals of desynchronization through dynamical indicators in systems undergoing saddle-node of cycles bifurcations.

Findings

Desynchronization in biological systems can be detected before full loss of resilience.

Traditional indicators fail for saddle-node of cycles bifurcations, but the proposed method succeeds.

Validated the approach using sleep-wake cycle models.

Abstract

Conditions such as insomnia, cardiac arrhythmia and jet-lag share a common feature: they are all related to the ability of biological systems to synchronize with the day-night cycle. When organisms lose resilience, this ability of synchronizing can become weaker till they eventually become desynchronized in a state of malfunctioning or sickness. It would be useful to measure this loss of resilience before the full desynchronization takes place. Several dynamical indicators of resilience (DIORs) have been proposed to account for the loss of resilience of a dynamical system. The performance of these indicators depends on the underlying mechanism of the critical transition, usually a saddle-node bifurcation. Before such bifurcation the recovery rate from perturbations of the system becomes slower, a mechanism known as critical slowing down. Here we show that, for a wide class of biological systems, desynchronization happens through another bifurcation, namely the saddle-node of cycles, for which critical slowing down cannot be directly detected. Such a bifurcation represents a system transitioning from synchronized (phase locked) to a desynchronized state, or vice versa. We show that after an appropriate transformation we can also detect this bifurcation using dynamical indicators of resilience. We test this method with data generated by models of sleep-wake cycles.