Hypothesis-Driven Specialization of Gene Expression Association Rules

This paper focuses on analyzing patterns mined from gene expression data. Whether a particular gene is "turned on" (expressed) or not is controlled by particular DNA se- quences (motifs). Multiple motifs are commonly involved in the expression of each gene, and the position and spacing of these motifs may be important. However, most available computational tools consider the importance only of indi- vidual motifs. We designed and developed an interactive tool that uses genetic data to derive association rules in- volving multiple motifs of possible significance in gene ex- pression. Genetic data are visualized in the context of a rule to facilitate rule specialization according to biological hypotheses regarding order, position, and spacing of mo- tifs. Different measures of interestingness (confidence, lift, p-value) are used to evaluate the rules\´ significance.