Consistent detection of cancer biomarkers with linear models

High throughput biological experiments such as DNA Microarrays are very powerful tools to understand and characterize multiple illnesses. These types of experiments, however, have also been described as large, complex, expensive and hard to analyze. For these reasons, analyses with linear assumptions are frequently bypassed for more sophisticated procedures with higher complexity. In this work, a search procedure for potential biomarkers using data from microarray experiments is proposed under purely linear assumptions. The method shows a high discrimination rate and does not require the adjustment of parameters by the user, thus preserving analysis objectivity and repeatability. A case study in the identification of potential biomarkers for cervix cancer is presented to illustrate the application of the proposed procedure.