CANCER CLASS PREDICTION: AN EMPIRICAL ANALYSIS OF GENE EXPRESSION DATA

摘要

We describe a method for binary cancer classificationrnbased on gene expression data from DNA microarrayrnhybridization experiments. The underlying architecture ofrnthis method is the gaussian basis function model but ourrnmodel development process is substantially different fromrnand superior to the current approaches. Cancer classrnprediction is crucial to its treatment and developing anrnanalytical approach for classification based upon thernmicroarray expression is an important problem. A genericrnapproach to classifying two types of acute leukemia basedrnon gene expression monitoring by DNA micro arrays wasrnoriginally pioneered by Golub et.al. [1]. In this paper, wernemploy their data sets to illustrate the use of our method,rnprovide insights into the classifier development processrnand undertake sensitivity analyses for selecting anrnappropriate cancer class prediction model.