Developing a novel test to detect cancer genes from microarray data

DNA microarray technology can simultaneously screen thousands of gene expression profiles, transforming how genetics is applied in medicine. However, the lack of normality in microarray data renders common statistical methods ineffective. We propose a novel statistical method which does not require stringent assumptions but is still more powerful than some of its competitors. Using both simulation studies and clinical data, we show that our novel method outperforms previous methods. The limiting distribution for the proposed test is obtained for under null and alternative hypotheses. The proposed test will help make cancer treatment and gene therapy more successful, and it may facilitate research regarding cancer vaccinations. The proposed test may also help in the development of a prediction model in genetic profiling studies built on a subset of differentially expressed genes and the clinical data to assess the accuracy of the clinical prediction.