The article studies the application of seven different methods of bioinformatics data replacement, whose working principles differ from each other, in real bioinformatics data classification. The practical experiments used three real bioinformatics data sets – Metabolic syndrome (Mets) patients – healthy obese subjects; Gastric cancer – healthy donors and Gastric intestinal disease – healthy donors. The missing data replacement was carried out using all studied methods and the classification data were compared. The paper also holds conclusions about the most efficient methods and the necessity to use such methods from the biologists’ point of view taking into account the real characteristics of data. The most important practical gain of the study is the biological analysis of If-Then rules produced by the applied classification algorithm FURIA, where the important conclusion is the ability of plasma leptin concentration to predict the risk of developing the Mets in a sample from an adult male population.