predicting the categories of colon cancer using microarray data and nearest shrunken centroid

Background & aim: it is very helpful to classify and predict the clinical category of a sample based on its gene expression profile. this study was conducted to predict tissues of colorectal adenoma, denocarcinoma, and paired normal in colon based on microarray data using nearest shrunken centroid method. methods & materials: in this study, the colon cancer dataset were used including, 18 adenocarcinoma, 4 colorectal adenoma, and 22 paired normal colon samples with 2360 common gene expression measurements. in order to predict categories of colon cancer was used nearest shrunken centroid method. r software was used for data analysis. results: based on our findings, performance of nearest shrunken centroid method was successful to reduce 2360 genes to a set of eleven genes containing rig, bigh3, gli3, homo sapiens guanylin, p78, 54kda, xbp-1, co-029, desmin, mlc-2, and hmg-1. this method predicted three classes. it predicted two classes-colorectal adenoma and adenocarcinoma with error of zero and normal class with error of 4.5%. conclusion: nearest shrunken centroid method succeeded to reduce several 1000 genes to 11 genes that were able to characterize colon tissue samples to one of the three classes of adenocarcinoma, colorectal adenoma and normal with 97.7% accuracy.