Optimal Experimental Design to Estimate Statistically Significant Periods of Oscillations in Time Course Data

We investigated commonly used methods (Autocorrelation, Enright, and Discrete Fourier Transform) to estimate the periodicity of oscillatory data and determine which method most accurately estimated periods while being least vulnerable to the presence of noise. Both simulated and experimental data were used in the analysis performed. We determined the significance of calculated periods by applying these methods to several random permutations of the data and then calculating the probability of obtaining the period's peak in the corresponding periodograms. Our analysis suggests that the Enright method is the most accurate for estimating the period of oscillatory data. We further show that to accurately estimate the period of oscillatory data, it is necessary that at least five cycles of data are sampled, using at least four data points per cycle. These results suggest that the Enright method should be more widely applied in order to improve the analysis of oscillatory data.