J. M. Sánchez Santos, F. J. Campos Laborie, J. De Las Rivas
Omic data from clinical studies are composed of quantitative measurements from biological features of a large number of factors, defining a complex phenotypic scenario. To characterize each categorical variable with continuous variables, multiple statistical approaches have been used in bioinformatics (PCA, DA, MANOVA, clustering). Here, several issues emerge for a correct characterization: based on parametric methods; heterogeneous omic data; focused on the similarity of samples instead of categories, etc. We propose a statistic to measure the proximity of the samples belonging to a category within a quantitative variable: given a quantitative variable whose values are ranked and assigned to categories of a factor, gaps are defined as distances between consecutive values belonging to each category. We determined the probability distribution of gaps for finding the most stable features, which will be those enclosing lower gaps among samples for each category of a factor.
Palabras clave / Keywords: bioinformatics, non-parametric feature selection, proximity measurement, stable patterns.
Programado
Pósteres I
30 de mayo de 2018 15:30
Zona EXPO
