Peano Count Trees (P-Trees) and Rule Association Mining for Gene Expression Profiling of Microarray Data

Abstract

The greatest challenge in maximizing the use of gene expression data is to develop new computational tools capable of interconnecting and interpreting the results from different organisms and experimental settings. We propose an integrative and comprehensive approach including a super-chip containing data from microarray experiments collected on different species subjected to hypoxic and anoxic stress. A data mining technology called Peano count tree (P-trees) is used to represent genomic data in multidimensions. Each microarray spot is presented as a pixel with its corresponding red/green intensity feature bands. Each bad is stored separately in a reorganized 8-separate (bSQ) file format. Each bSQ is converted to a quadrant base tree structure (P-tree) from which a superchip is represented as expression P-trees (EP-trees) and repression P-trees (RP-trees). The use of association rule mining is proposed to derived to meanigingfully organize signal transduction pathways taking in consideration evolutionary considerations. We argue that the genetic constitution of an organism (K) can be represented by the total number of genes belonging to two groups. The group X constitutes genes (X1,Xn) and they can be represented as 1 or 0 depending on whether the gene was expressed or not. The second group of Y genes (Y1,Yn) is expressed at different levels. These genes have a very high repression, high expression, very repressed or highly repressed. However, many genes of the group Y are specie specific and modulated by the products and combinations of genes of the group X. In this paper, we introduce the dSQ and P-tree technology; the biological implications of association rule mining using X and Y gene groups and some advances in the integration of this information using the BRAIN architecture.