Using perceptrons for supervised classification of DNA microarray samples: Obtaining the optimal level of information and finding differentially expressed genes

Abstract

The success of the application of neural networks to DNA microarray data comes from their efficiency in dealing with noisy data. Here we describe a combined approach that provides, at the same time, an accurate classification of samples in DNA microarray gene expression experiments (different cancer cell lines, in this case) and allows the extraction of the gene, or clusters of coexpressing genes, that account for these differences. Firstly we reduce the dataset of gene expression profiles to a number of non-redundant clusters of coexpressing genes. Then, the cluster's average values are used for training a perceptron, that produces an accurate classification of different classes of cell lines. The weights that connect the gene clusters to the cell lines are used to asses the relative importance of the genes in the definition of these classes. Finally, the biological role for these groups of genes is discussed. © Springer-Verlag Berlin Heidelberg 2002.