Albeit the great number of microarray data available on breast cancer, reliable identification of genes associated with breast cancer development remains a challenge. The aim of this work was to develop a novel method of meta-analysis for the identification of differentially expressed genes integrating results of several independent microarray experiments. We developed a statistical method for identification of up- and down-regulated genes to perform meta-analysis. The method takes advantage of hypergeometric and binomial distributions. Using our method we performed meta-analysis of five data sets from independent cDNA-microarray experiments on breast cancer. The meta-analysis revealed that 3.2% and 2.8% of the 24,726 analyzed genes are significantly (P-value < 0.01) down- and up-regulated, respectively. We also show that properly applied meta-analysis is a good tool for comparison of different breast cancer subtypes. Our meta-analysis showed that the expression of the majority of genes does not show significant differences in different subtypes of breast cancer. Here, we report the rationale, development and application of meta-analysis that enable us to identify biologically meaningful features of breast cancer. The algorithm we propose for the meta-analysis can reveal the features specific to the breast cancer subtypes and those common to breast cancer. The results allow us to revise the previously generated lists of genes associated with breast cancer and also identify most promising anticancer drug-target genes.