A normalization-robust bootstrap-based rhythmicity measure to detect rhythmic genes in oscillatory systems

Abstract

Microarray gene expression data are extremely noisy. Normalization is widely regarded as an essential step before data analysis to remove systematic variations while maintaining biological signals of interest. However, the choice of normalization may substantially impact the detection of rhythmic genes in oscillatory systems. We introduce a rhythmicity measure and a bootstrap methodology to detect rhythmic genes robust with respect to the normalization choice.