Statistical resynchronization and detection of periodic transcripts

A periodic-normal mixture (PNM) model is proposed to fit the yeast cell cycle data of Cho et al. and Spellmen et al. and to resynchronize the transcription profiles of cell-cycle related genes. Subsequently a two-component beta mixture model is used to approximate the distribution of the PNM model fitting residues so as to compute the posterior probability for each gene to be cell-cycle related. Our result suggested that ∼32% genes in yeast genome are likely to be periodically expressed, among which 822 genes met our stringent criterion. Of the 822 genes, 282 are absent from the list of 800 genes reported by Spellman et al., including many known important cell-cycle related functional genes. Phase matching of the 822 resynchronized expression profiles implied that the three synchronization methods might have brought cells to the same phase at the time of release.