Support Vector Machines for Biomarkers Detection in in vitro and in vivo Experiments of Organochlorines Exposure

Metabolomic studies generate large amounts of data, whose complexity increases if they are derived from in vivo experiments. As a result, analysis methods highly used metabolomics, such as Partial Least Squares Discriminant Analysis (PLS-DA), can have particular difficulties with this type data. However, there is evidence that indicates Support Vector Machines (SVMs) better deal complex On the other hand, chronic exposure to organochlorines public health problem. It has been associated diseases cancer. Therefore, its identification relevant reduce their impact on human health. This study explores performance SVMs classifying metabolic profiles and identifying metabolites organochlorines. For purpose, two experiments were conducted: first one, organochlorine was evaluated HepG2 cells; and, second it serum samples agricultural workers exposed pesticides. The compared PLS-DA. Four kernel functions assessed SVMs, accuracy both using k-fold cross-validation test. In order identify most metabolites, Recursive Feature Elimination (RFE) Variable Importance Projection (VIP) results show exhibit higher percentage fewer training workers. Finally, workflow based for biomarkers high biological proposed.