Aim.To determine the expression profiles of a set of cancer-associated genes in prostate tumors, using various normalization protocols (with 1, 2 and 4 reference genes) and to optimize a combination of reference genes to calculate the relative expression (RE) of the investigated genes in prostate cancers. Methods. Relative expression level of 23 genes was analyzed by quantitative PCR (qPCR) in 37 prostate cancer tissues (T) with different Gleason scores (GL) and at various stages and compared with 37 corresponding normal prostate tissue (CNT) samples and with 20 samples of prostate adenomas. Results. Theoretical calculations of the RE deviation showed no influence of the normalization protocols on the results for both the reference and the investigated genes. The experimental data that were calculated using a 2 -ΔΔCt showed statistically significant differences in the expression of 17 out of 23 investigated genes, when the paired T/CNT were compared. RE values calculated using the 2 -ΔCt method showed a high similarity of statistical data in all reference gene groups for tumor-CNT-adenoma groups (> 82 %). Data grouping by a cancer stage showed 69 %, and by the GL score -64.5 % of the data overlapping. Conclusions. All three types of normalization protocols, as expected, can be used for RE normalization in prostate tumor samples. The usage of either the 2 -ΔCt or 2 -ΔΔCt models showed no difference in the calculated RE levels for the studied reference genes. The most important factor was the constitutive expression of the reference genes. Moreover, the expression levels of the investigated genes, changes in RE values, number of samples in groups and heterogeneity of gene expression are important parameters for the selection of the threshold in expression level differences between groups for a reliable data interpretation.K e y w o r d s: prostate tumors, relative expression, reference genes validation, expression level, genes expressed at low levels.