This study was a part of a FIRB-2001 collaborative project aiming at comparing gene transcription profiles in different swine tissues, to identify genes candidate to the control of trait differences between Large White and Casertana swine breeds. We sampled liver tissue from 17 Casertana and 16 Large White animals at slaughter and assayed them on a microarray comprising 10,665 selected 70 mer oligo probes representing Sus scrofa gene sequences with a hit to human, mouse and pig gene transcripts and some sequences containing a 3’ expressed sequence tag (EST). In this poster we focus on the comparison between alternative experimental designs for the production of gene expression data. Microarrays were hybridized with either individuals or pooled biological samples from different breeds. Each experiment was quadruplicated (dye-swap and complete replication). The different experimental designs planned were comparison: i) between single animals ii) between the same animals in pools of 3 individuals per breed iii) between independent pools of 5 individuals per breed. Loess normalization was applied within slide to normalize for intensity- and spatially-dependent systematic error. A mixed linear model was used for normalization across slides. The residuals from the normalization process were used to test for effects on gene expression. Differences in the least square means were obtained and tested for significant differences by using a t test. The false discovery rate at P≤0.05 was used to account for multiple testing. Repeatability of the different experimental designs were calculated and the top lists of differentially expressed genes compared. Repeatability was approximately 60% for all the experimental designs, yielding slightly higher values for comparison between individuals. Most of the genes ranking highest for significance were common to the top list for the different designs indicating that, compared to the analysis of single individuals, pool analysis reduces the number of comparisons and cost without significantly altering the final results.