[From biomarkers to clinical decision support].

The rapidly accumulating, heterogeneous, numerous biomedical data and the increasing knowledge represent both the key areas and the major bottleneck in the complete realization of personalized medicine. Traditional clinical aims as well as decision-making processes in drug research need systems-based integration of data and their analyses, which require a wide range of new mathematical approaches and application of information technologies. In the world of biomarkers, all such tasks correspond to three types of biomarkers, namely, endpoint, target and diagnostic biomarkers, which together form a complex network. To decipher the networks, probabilistic graphical models are introduced. Briefly, the authors illustrate their use for exploration and visualization of direct dependencies among biomarkers. Finally, construction of decision networks and their application to structural and quantitative optimization of biomarkers are discussed to provide a novel type of supporting tools for clinical practice.