Analysis Of An Associative Memory Neural Network For Pattern Identification In Gene Expression Data
Silvio Bicciato, Mario Pandin, Carlo Di Bello · 2001
DNA microarrays are becoming a standard tool for determining the role of genes in the regulation of any biological process in an organism. The application of this technology for the analysis of gene expression creates enormous opportunities for accelerating the pace towards the understanding of living systems and for the identification of target genes and pathways for drug development. However, equally efficient methods need to be developed for upgrading the information content of the large amounts of data generated by microarray experiments. A procedure for extracting patterns of gene expression through the analysis of the architecture of an associative memory neural network is described. Such patterns contain critical information about the gene-networking relationships observed during changes in cell physiology and the onset of diseases. The proposed method has been tested on two different microarray data sets, namely DeRisi's experiment on yeast cultures [10] and Golub's analysis of acute human leukemia molecular profiles [17]. Using these data sets, the neural network structure has been examined to extract relationships among different genes involved in major metabolic pathways and to relate specific genes to different classes of leukemia. Keywords Gene expression data, cDNA microarrays, neural networks, pattern recognition, data mining 1.