An Evaluation of Gene Ranking Methods by Gene Ontology

In this paper, we use three gene ranking techniques including two new methods based on a single gene score approach and support vector machine (SVM) method based on recursive feature elimination (RFE). These methods have been evaluated on colon cancer dataset by gene ontology. Experimental results showed that these methods select genes that do not have any GO. So, great parts of data that have GO are deposited. With having GO, we can gain a better understanding of the data than just applying statistical analysis because biological significance does not necessarily have to be statistically significant. We propose to enter biological knowledge into gene selection process. Using GO prevents from producing spurious results

1. Introduction

The DNA microarray technology is emerging recently in the field of computational biology. This technology enables monitoring and measurement of expression levels for thousand of genes simultaneously through the hybridization process [1]. It is commonly used for comparing the gene expression levels in tissues under different conditions, such as wild-type versus mutant, or healthy versus diseased [2]. A typical DNA microarray study utilizes several DNA microarray chips on different tissue samples and generates a numerical array with thousands of rows (genes) and tens of columns (experiments/DNAchips) [4]. The gene expression indicates the approximate amount of protein contents in a given sample [1]. Microarray can be applied to a wide range of studies including gene regulation, disease diagnosis and prognosis, cancer classification, bio-marker discovery and drug development [5].