A Maximal Eigenvalue Method for Detecting Process Representative Genes by Integrating Data from Multiple Sources

Abstract

An important problem in computational biology is the identification of candidate genes which can be considered as representative of the different cellular processes taking place in the cell as it evolves through time. Multiple and very noisy data sources contain information about such processes and should therefore be inte-grated in order to obtain a reliable identification of such candidate genes. In this paper, we present a novel ranking algorithm which determines process represen-tative genes by integrating a set of noisy binary relations between genes. We present some preliminary results on two artificial toy datasets and one real biolog-ical problem. In the biological problem, we use this method to identify representa-tive genes of some of the fundamental biological mechanisms taking place during cellular growth in A. thaliana by integrating gene expression data and information from the gene GO annotation.