Data Analysis of Arabidopsis Tiling Array

DNA tiling microarray technology has become a major bioinformatics tool for genomic research. Due to the high-density, high-throughput characteristics, tiling array can help to study gene expression and to explore the mystery of life from genome level. However, due to its data volume and complexity, the analysis of tiling array data is not streamlined yet. Although some dynamic programming approaches have been successfully applied to yeast tiling array data, the segmentation problem is considerably more challenging for the genomes of higher eukaryotes, such as Arabidopsis. In this paper, we applied a new machine learning method combining the advantages of Hidden Markov (HM) models and Support Vector Machines (SVM) to deal with the Arabidopsis tiling array data by adopting the probe filtering and normalization of wild type samples to identify gene structures.