Interactive Data Mining Tool for Microarray Data Analysis Using Formal Concept Analysis

The DNA microarray analysis can explain gene functions by measuring tens of thousands of gene expressions at once and analyzing gene expression profiles that are obtained from the measurement. However, gene expression profiles have such a vast amount of information and therefore most analyses work are done on the data narrowed down by statistical methods, there remains a possibility ofmissing out on genes that consist the factors of phenomena from their evaluations. This study propose a method based on a formal concept analysis to visualize all gene expression profiles and characteristic information that can be obtained from annotation information of each gene so that the user can overview them. In the formal concept analysis, a lattice structure that allows genes to be hierarchically classified and made viewable is built based on the inclusion relations of attributes from a context table in which gene is the object and the attributes are expression profiles and binarized characteristic information. With the proposed method, the user can change the overview state by adjusting the expression ratio and the binary state of characteristic information, understand the relational structure of gene expressions, and carry out analyses of gene functions. We develop software to practice the proposed method, and then ask a biologist to evaluate effectiveness of proposed method applied to a function analysis of genes related to blue light signaling of rice seedlings.