Sánchez Peña, Matilde L.

Profile Picture

Filters

20101
Reset filters

Settings

Citations

Publication Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    Identification of potential cancer biomarkers through multiple criteria optimization using microarray data
    (2010) Sánchez Peña, Matilde L.; Cabrera Ríos, Mauricio; College of Engineering; Castro, José M.; Medina Borja, Alexandra; Isaza Brando, Clara E.; Department of Industrial Engineering; Sharma, Anand D.
    Cancer is a worldwide relevant illness given its mortality rates and associated economic and social repercussions. Genetic profiling has become one of the most important tools for cancer characterization, its diagnosis and prognosis. Microarrays are biological experiments that have been used in recent years with this end in mind due to their capacity to measure the relative genetic expression of tens of thousands of genes simultaneously. One of the principal aims using data from microarray experiments is the selection of relevant genes that can be used as surrogate measures for the state of cancer, i.e. cancer biomarker genes. Many and varied methodologies have been developed and used for this purpose ranging from the simplest statistical approaches to sophisticated Artificial Intelligence methods. The explored literature, however, shows that setting parameters for several of these approaches is often a difficult task for final users, who mainly hail from the biological and medical sciences. As a consequence, analysis results have been reported to vary across different researchers even when using the same microarray datasets. This situation is an opportunity to develop methodologies to find potential cancer biomarkers in a consistent manner. In this work potential biomarker identification is casted as a Multiple Criteria Optimization (MCO) Problem, aiming to remove analysis subjectivity due to parameter adjustment. MCO is a methodology used to find the best compromises between two or more conflicting criteria.The main proposition of this work is that several measures related to microarray data analysis can be seen as criteria to be optimized. It is desirable, for example, that the p-value associated to a particular gene be low when trying to determine its statistical significance. If a gene could be characterized through two or more p-values, then an MCO problem can be formulated. Solving an MCO problem results in a set of solutions representing the best compromises among all the considered criteria. These solutions are called Pareto-efficient solutions and they conform a so-called efficient frontier of the problem. This work proposes that genes on the resulting efficient frontier of an associated MCO problem could be cancer biomarkers. Among the methodologies used to solve MCO problems, Data Envelopment Analysis (DEA) has been chosen in this work because it does not require parameter setting by the user in many of its possible formulations. Furthermore, DEA can be solved through linear programming, the most tractable of optimization problems and for which inexpensive commercial software readily available. To the best extent of our knowledge, this work constitutes the first effort on using Multiple Criteria Optimization to detect potential cancer biomarkers from microarray data.