GNN - Genome News Network  
  Home | About | Topics
Classifying cancers with DNA microarrays and artificial neural networks
By Birgit Hofmann-Reinert

Researchers have developed a new method for classifying four seemingly similar types of childhood cancers based on their gene-expression data. By using DNA microarrays combined with artificial neural networks, the research team diagnosed the different cancers with a high degree of accuracy.

Under the microscope, the cells of the four cancers—neuroblastoma, rhabdomyosarcoma, non-Hodgkin lymphoma, and the Ewing family of tumors—look very much alike and can therefore be hard to diagnose. To tell them apart, Paul S. Meltzer, of the Cancer Genetics Branch at the National Institutes of Health, in Bethesda, Maryland, and colleagues studied the expression levels of the genes involved.

More than 6,000 genes were simultaneously analyzed using DNA microarrays. The large amount of data was then 'fed' into an artificial neural network (ANN). A neural network is an optimization method, which models the behavior of neurons in the brain. It can be trained to recognize and categorize complex patterns.

"To our knowledge, this is the first application of ANN for diagnostic classification of cancer using gene-expression data derived from cDNA microarrays," the researchers write in the June issue of Nature Medicine. After the neural network had learned to recognize the genetic pattern of each cancer type, it correctly classified 20 test samples into four subgroups respectively. It also correctly excluded five samples of other types of cancer and healthy cells.

"Future applications of these methods will include studies to classify cancers according to stage and biological behavior in order to predict prognosis and thereby direct therapy," the researchers conclude.

. . .

Khan, J. et al. Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks. Nat Med 7, 673-679 (June 2001).

Back to GNN Home Page