Abstract
An emergence of tiling arrays enables to make a comprehensive and a quantitative gene analyses. Toyoda and Sinozaki developed a genome wide gene prediction method, called ARTADE (ARabidopsis Tiling-Array-based Detection of Exons). The method had succeeded to detect numerous unknown non-coding genes from Arabidopsis. However, the method uses tiling arrays information independently for each probe. The conventional ARTADE is therefore weak for noisy probes. Moreover, if two genes are closed each other, it is difficult to divide the two genes from redundant tiling array sequence. For the problems, we have developed the new ARTADE(ARTADE2.0) based on multiple tiling array data. The method predicts genes by spatial probabilistic model based on correlation matrix score of multi-dimensional data by tilling arrays. ARTADE2.0 shows robustness for the noisy probes and a high performance for the start-end prediction of genes.
