Machine learning approaches to identify sleep genes

Abstract

Abstract Genetics impacts sleep, yet, the molecular mechanisms underlying sleep regulation remain elusive. We built machine learning (ML) models to predict genes based on their similarity to known sleep genes. Using a manually curated list of 109 labeled sleep genes, we trained a prediction model on thousands of published datasets, representing circadian, immune, sleep deprivation, and many other processes. Our predictions fit with prior knowledge of sleep regulation and also identify several key genes/pathways to pursue in follow-up studies. We tested one of our findings, the NF-κB pathway, and showed that its genetic alteration affects sleep duration in mice. Our study highlights the power of ML to integrate prior knowledge and genome-wide data to study genetic regulation of sleep and other complex behaviors.