12345678910 Essay

7414 WordsJun 3, 201430 Pages
0 6 Metabolic Biomarker Identification with Few Samples Pietro Franceschi, Urska Vrhovsek, Fulvio Mattivi and Ron Wehrens IASMA Research and Innovation Centre Via E. Mach, 1 38010 S. Michele all’Adige (TN) Italy 1. Introduction Biomarker selection represents a key step in bioinformatic data processing pipelines; examples range from DNA microarrays (Tusher et al., 2001; Yousef et al., 2009) to proteomics (Araki et al., 2010; Oh et al., 2011) to metabolomics (Chadeau-Hyam et al., 2010). Meaningful biological interpretation is greatly aided by identification of a “short-list” of features – biomarkers – characterizing the main differences between several states in a biological system. In a two-class setting the biomarkers are those variables (metabolites, proteins, genes ...) that allow discrimination between the classes. A class or group tag can be used to distinguish many situations: it can be used to discriminate between treated and non-treated samples, to mark different varieties of the same organism, etcetera. In the following, we will – for clarity – restrict the discussion to metabolomics, and the variables will constitute concentration levels of metabolites, but similar arguments hold mutatis mutandis for other -omics sciences, such as proteomics and transcriptomics, where the variables correspond to protein levels or expression levels, respectively. There are several reasons why the selection of biomarker short-lists can be beneficial: • Predictive purposes: using only a small number of biomarkers in predictive class modeling in general leads to better, i.e., more robust and more accurate predictions. • Interpretative purposes: it makes sense to first concentrate on those metabolites that show clear differences in levels in the different classes, since our knowledge of metabolic networks in many cases is only scratching the surface. • Discovery purposes: the

More about 12345678910 Essay

Open Document