Abstract
A simple, nonparametric and distribution free method was developed for quick identification of the most meaningful biomarkers among a number of candidates in complex biological phenomena, especially in relatively small samples. This method is independent of rigid model forms or other link functions. It may be applied both to metric and non-metric data as well as to independent or matched parallel samples. With this method identification of the most relevant biomarkers is not based on inferential methods; therefore, its application does not require corrections of the level of significance, even in cases of thousands of variables. Hence, the introduced method is appropriate to analyze and evaluate data of complex investigations in clinical and pre-clinical basic research, such as gene or protein expressions, phenotype-genotype associations in case-control studies on the basis of thousands of genes and SNPs (single nucleotide polymorphism), search of prevalence in sleep EEG-Data, functional magnetic resonance imaging (fMRI) or others.