47glioblastoma Gene Expression Profile Diagnostics by the Artificial Neural Networks1

Abstract

Two artificial neural networks of different types were applied to profiles of gene expression in most aggressive human brain tumor – glioblastoma – and in normal brain tissue. The results of gene expression profiles classification are presented. First method – self organizing maps – gave good dis crimination of profiles on the trained map. Another ANN – perceptron – showed a good result of classification – more then 95% of the test data set were successfully classified. Due to high correlations between some gene expression values one can suppose, that number of genes necessary for successful classification may be reduced.