(341a) Identifying Circadian Drug Targets for Maintained Oscillatory Precision

The mammalian circadian clock is a complex dynamic system comprised of interlocked transcription-translation feedback loops. Approximately 10% of all mammalian transcripts are under circadian regulation, highlighting the importance of this system. Disruptions such as shift work, high-fat diets, or aging have been shown to reduce the amplitude of circadian oscillation, and these diminished amplitudes have been associated with a higher incidence of metabolic disorders such as diabetes and obesity. There has therefore been significant recent interest in identifying pharmaceutical modulators of the clock to support robust circadian function. The cellular circadian oscillator is subject to intrinsic molecular fluctuations caused by the low copy number of biomolecules, resulting in cycle-to-cycle variability in period length for cells in the absence of synchronizing cues. Recent work has shown that potential therapeutic treatments affect the precision of the clock, in addition to the period, phase, or amplitude. Furthermore, the vast majority of clock hits within a recent genome-wide siRNA screen resulted in either decreased oscillatory amplitudes or decreased clock precision, indicating that identifying suitable drug targets in the circadian network is a significant challenge.

Rational design of circadian modulators is desirable; however, computational screening for changes in oscillator precision generally requires discrete stochastic simulation of hundreds or thousands of trajectories and then fitting a phase diffusion coefficient for each possible drug target. This approach is prohibitively computationally expensive. Here, we present an efficient analytical method for determining the precision of the circadian oscillator subject to intrinsic molecular noise. Briefly, we construct a model-reduced oscillator with relaxation dynamics governed by the Floquet multipliers of the system. By calculating the phase sensitivity to perturbation and approximating the noise distribution, we are able to analytically calculate a phase diffusion coefficient. We demonstrate agreement between this method and full stochastic simulation. Finally, we apply this method to a computational model of the circadian clock, and identify potential therapeutic targets which allow manipulation of clock period without diminishing precision.