Abstract
For a firm gene annotation, it is necessary to repeat experiments and obtain a robust result. With more stack observations of transcripts, the clearer gene structure would be seen. Then, we developed a statistical genome-wide gene prediction method ''ARTADE2.0'' from the discovery that the correlations matrix made with tiling array probes of many experiments shows transcription units clearly. However, the estimation result may be mixed gene model if an alternative splicing (AS) model exists since expressions of transcripts are observed from plural kinds of RNAs. We now proposed a method for a detection of AS part based on a factor analysis (as a statistical methods). The method can divide the gene region into part belonging to major AS isoform and part belonging to minor AS isoform or alternatively spliced region. For the method, we used tiling arrays results from eighteen conditions, including plants under various stresses or several organs. The method succeeded to elucidate some known ASs and other novel alternatively spliced regions in genes extracted through the ARTADE2.0.
