Cluster-based characterization of gene over-expression in cancer sets

Mining techniques are needed to extract important information from huge high dimensional gene expression sets. Targeting unique expression behavior as over/under-expression is specific to gene expression data and is needed to explore another direction in the relation of genes to tumor conditions. This research proposes criteria for filtering over-expression genes, identifying over-expression related samples and using them to characterize over-expression behaviour in gene clusters and outliers. In return, hypothetical marker genes and functional relations can be provided, ready for approval by the aid of other datasets/results. Experiments are performed on breast cancer expression data.