Our work consists in developing a workflow using Knowledge Discovery methodologies to propose advanced predictive biomarkers discovery solutions from metabolomic data. We propose to use machine learning algorithms for feature selection and FCA for visualization.