The research on biomarkers to detect livestock treated with recombinant bovine somatotropin (rbST) is still an open issue. In fact, beyond undertaking confirmation methods, there is the need to develop simple and inexpensive screening tests. In this direction, some proposals have been forwarded, mostly involving the measurement of circulating molecules, whereas the possibility of using biomarkers related to gene expression is a field under investigation. The present study was carried out on sixteen buffalos, eight of which treated with rbST. Blood samples were collected six times during the treatment to investigate on the presence of differentially expressed genes in leukocytes. Analysis with the microarray technique was performed on two sampling moments, in order to obtain a first selection of genes. Further analysis was carried out by real time RT-PCR, in order to create a discriminating linear system. A study on the variation of the error related to the number of samples included in statistics was also performed. Results showed that, including an increasing number of samples to build the discriminating algorithm, the-error grows and tends to stabilize on 6.5%. This study clearly shows the paramount importance of including a proper number of samples to obtain reliable algorithms.