This application proposes to develop methodology to quantify sleep and its substages using high throughput technology based on digital analysis of videos as a screening tool for altered sleep and wake. There are many disorders that affect sleep. These can result in sleep disturbance with excessive sleepiness during the day. Also, complaints of sleeping difficulty, i.e., insomnia, are extremely common. The high prevalence of these disorders has resulted in a major interest from some of the largest pharmaceutical companies to develop improved medications for these conditions. An important barrier is that currently sleep can only be assessed in animal models (mice and rats) used for drug development by recording brainwaves (EEG) and muscle activity (EMG). This requires anesthesia of the animal, surgery with implantation of electrodes, and time (days) to recover from surgery before the animal is studied. This technology is not particularly suited for high throughput screening of the effect of the large number of compounds that is required for drug development. High throughput screening is also needed to study the large number of knockout mice that are increasingly available. Determining which genes affect sleep is another strategy for drug discovery. Our group has recently developed a novel approach based on study of movement and quiescence from analysis of videos. We have shown that we can reliably and accurately identify and quantify sleep. But it is not sufficient to estimate sleep since sleep contains different substages, i.e., rapid eye movement (REM) sleep and non-rapid- eye movement (NREM) sleep. Different drugs and genes can differentially affect these different states. Thus, a high throughput screening strategy needs to be able to also identify and quantify the substages of sleep. Our preliminary work indicates that there are features of video analysis that will allow us to separately quantify NREM and REM sleep. To do so, we propose to utilize techniques from statistics/mathematics applying ensemble-learning algorithms to identify patterns in large multivariate datasets, i.e., different variables in a time series obtained from video analysis, for the purpose of classification and prediction. This is, we propose, the first step in commercialization of this new technology which has many applications. The application represents a collaboration between the University of Pennsylvania and Neurocare, two organizations that have a successful history of collaborating on projects. This application is to develop a technique to simply assess, in mice, the amounts of sleep, wake and the different stages of sleep, including dream sleep. The approach is based on careful examination of continuous pictures of the mouse over many hours. The new technique will allow screening of large numbers of mice to determine drugs that likely alter sleep and large numbers of mice with altered function of a particular gene. ? ? ?
| McShane, Blakeley B; Jensen, Shane T; Pack, Allan I et al. (2013) Statistical Learning with Time Series Dependence: An Application to Scoring Sleep in Mice. J Am Stat Assoc 108:1147-1162 |
| McShane, Blakeley B; Galante, Raymond J; Biber, Michael et al. (2012) Assessing REM sleep in mice using video data. Sleep 35:433-42 |
