A probabilistic model of sleep structure suggests a primary role for wake probability in overall sleep in Drosophila melanogaster

In humans and flies, sleep is a highly organized behavior, occurring in discrete bouts that are consolidated at night but have a more fragmented structure during the day. Longer episodes correlate with transitions to deeper sleep states for both species, and sleep fragmentation has been shown to have deleterious outcomes for a large number of physiological processes. In mammals and birds, EEG provides an excellent measure of sleep structure and state. In invertebrates, assessment of sleep structure has relied on measures of episode duration, which can fail to capture important aspects of sleep architecture. Here we model Drosophila sleep as a probabilistic behavior and develop analytical tools to measure transition probabilities and the organization of sleep episodes within a defined analysis window. We show that these tools provide insight into the underlying mechanisms of normal sleep and into how they are altered in pathological conditions. The probability of transition to the wake state, P(Wake) is a major driver of normal sleep and responsive to dopaminergic tone, suggesting it is a measure of arousal. Our method for assessing sleep fragmentation adds important information about the organization of sleep bouts that is missing in measures of episode duration alone. Importantly, the ability to obtain a measure of the probability structure underlying sleep in a particular condition or genotype provides a way to compare the state of brain sleep circuits across conditions that utilizes information from all of the many processes that influence sleep.