0038 Inferring the Genetic Basis of Sleep States in Drosophila with Hidden Markov Models (HMMs)

Abstract Introduction Sleep is essential, yet its genetic regulation is not fully understood. Drosophila melanogaster is as a valuable model for sleep research due to conserved genetic similarities with humans. Although traditionally viewed as having a single sleep state, recent evidence suggests flies experience multiple sleep states of varying depths, like mammals. Here, we classify fly sleep states from locomotor activity data and map the time spent in each state to the genome. Methods We used HMMs on activity data from 10,130 Drosophila flies across 168 genotypes. Models with 2 to 10 sleep/wake states were tested and the best one was selected using statistical and information theory criteria. We iterated the HMM multiple times to assess the accuracy of the model. We also used arousal threshold testing to confirm model accuracy. We performed Genome Wide Association (GWA) with 2,189,692 polymorphisms to find candidate genes associated to each sleep/wake state. Results A 4-state HMM with 1000 iterations achieved over 99% accuracy in classifying sleep/wake states. We defined the states as State0 (active wake), State1 (quiescent rest), State2 (light sleep), and State3 (deep sleep) based on activity level. Flies spent an average of 100.34 min in State0, 126.49 min in State1, 147.96 min in State2, and 1065.45 min in State3 per day. Deep sleep had the highest arousal threshold (0.52g), followed by light sleep (0.35g), quiescent rest (0.31g), and active wake (0.26g). Genetic analysis showed significant heritability (0.27-0.50) affecting sleep states. GWA identified 989 SNPs in 505 candidate genes, 376 of which have human orthologs. Network analyses suggested potential hub genes in sleep state regulation, including transcription factors, enzymes, neurotransmitters, extracellular matrix proteins, and neuronal remodeling proteins. Conclusion Our study shows that Drosophila has distinct, heritable lighter and deeper sleep states. We identified genetic variants common and unique to these states, suggesting targets for future research. This framework, available as an open-source R package, offers a powerful tool to explore genetic mechanisms of sleep. Support (if any) This work was funded by the Division of Intramural Research at the National Heart, Lung, and Blood Institute (NHLBI). AG is supplemented by a L’Enfant fellowship from NHLBI.