The analysis of complex and massive biological data issued from metabolomic analytical platforms is a challenge of high importance. The analyzed datasets are constituted of a limited set of individuals and a large set of features where predictive biomarkers of clinical outcomes should be mined. Accordingly, in this paper, we propose a new hybrid knowledge discovery approach for discovering meaningful predic-tive biological patterns. This hybrid approach combines numerical classifiers such as SVM, Random Forests (RF) and ANOVA, with a symbolic method, namely Formal Concept Analysis (FCA). The related experiments show how we can discover among the best potential predictive biomarkers of metabolic diseases thanks to specific combinations of clas-sifiers mainly involving RF and ANOVA. The visualization of predictive biomarkers is based on heatmaps while FCA is mainly used for visual-ization and interpretation purposes, complementing the computational power of numerical methods.