Mining microarray expression data for classifier gene-cores

摘要: We present algorithms and methods for mining DNA microarray expression data from a classification perspective. demonstrate these on gene dataset containing two acute leukemia classes. A prioritized feature-selection approach is followed to account incomplete knowledge of function complex inter-gene dependencies. utilize combination class scatter metrics heuristic search determine all those minimal combinations genes that have potential discriminate between the modified perceptron training algorithm further trains discriminant gene-sets. This process results in large number distinct accurate classifiers (gene sets). an which then employed mine discover 'core' patterns compositions classifier These cores are potentially very useful biologist may reveal much about dependencies functions.