Finding Clocks in Genes: a Bayesian Approach to Estimate Periodicity

• Activity Number: 338
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Yan Ren*+ and Christian Hong and Sookkyung Lim and Seongho Song
• Companies: and University of Cincinnati and University of Cincinnati and University of Cincinnati
• Keywords: Bayesian Analysis ; Bayes Factor ; Period Analysis ; Circadian Rhythms ; Ultradian Rhythms

Abstract

We successfully developed a Bayesian Analysis of Autoregressive Spectral Estimation Regression (ABSR) to identify patterns and assign classifications (Arrhythmic, Ultradian, and Circadian) of oscillating gene expression in the data. Currently, JTK_CYCLE (Hughes et al., 2010) is widely used to identify the periodicity of gene expression. At the same time, Yang and Su (2010) developed autoregressive spectral estimation regression (ARSER) for microarray data observed at sparse time-points. Based on our ABSR simulations, we highly improved True Discovery Rate (TDR) and reduced False Discovery Rate (FDR) compared with JTK_CYCLE. In comparison with the ARSER method, we reduced FDR by 5-18% with the same level of TDR (~90%). Furthermore, our simulation studies indicate that JTK_CYCLE is not consistent when the interested range of periods is different. In contrast, the ABSR method gives consistent estimation of periods. In the end of this article, we applied these models to mammalian cell line data (NIH3T3 and U2OS) provided by John Hogenesch's lab.