抄録
The recent developments in high-throughput technologies such as cDNA and oligo nucleotide microarray have enabled us to simultaneously analyze thousands of gene expressions. Gene clustering is one of the most important analyses for DNA microarray data. Many of the clustering methods require the whole expression time course profiles, however, most predominant information is time period of onset of each gene. In the gene network or circuit, gene switches for expression are turned on or off according to their demand. It should be effective to cluster the gene expression data based on expression profile considering both onset and cessation of each gene, however, there are few reports on such clustering methods. In this study, we estimated timing of gene expression onset and its cessation by simultaneous differential equation model. Since many of the values of parameters on mRNA transcription rate, mRNA decay rate, onset and cessation time of gene and cessation are all still unknown, we optimize these parameter-values for experimental data. Then we compare clustering result using estimated parameter-values and all expression time course profiles. In the next, we examined the influence of noise upon clustering result and evaluated the clustering result by Wilcoxon signed rank test. Clustering result shows that estimated parameter-values realized the almost same result as that of whole time course data. Furthermore, clustering method using estimated parameter-values was more noise tolerance than that using whole time course data. (p=0.005) These results suggest that, the clustering method using the estimated onset time and cessation is useful analysis method to cluster the experimental data.
