Classifying Gene Expression Patterns

Carten Peterson

Abstract
Supervised methods for analyzing microarray data are discussed;
signal-to-noise methods, support vector machines and multilayer
perceptrons
with principal component preprocessing. The methods are illustrated with
clinical applications for diagnostics and gene extraction; small round
blue
cell tumors (SRBCT) of childhood and breast cancer. In the context of
gene extraction, pathway "depths" are investigated. The importance of
using
blind test sets for evaluation is stressed. Also, sample selection
issues
are emphasized. Advantages of merging information from different kinds
of arrays are discussed.