Detecting Periodically Expression in Unevenly Spaced Microarray Time Series

Abstract

Abstract. Spectral estimation has important applications to microarray time series analysis. For unevenly sampled data, a common spectral estimation technique is to use the Lomb-Scargle algorithm. In this paper, we introduce a new reconstruction algorithm and singular spectrum analysis (SSA) method to deal with unevenly sampled microarray time series. The new reconstruction method is based on signal reconstruction technique in aliased shift-invariant signal spaces and a direct implemental algorithm is developed based on the B-spline basis. We experiments on simulated noisy signals and gene expression profiles show different effects for our designed three methods. The three methods are based on our presented reconstruction algorithm, SSA, classical FFT periodogram method and Lomb-Scargle periodogram method.