POSTERS ABSTRACTS 2 New Biotechnology · Volume 27S · April 2010 However, increasing evidence challenges the model that bipolar disorder and schizophrenia would be independent syndromes. We performed meta-analysis of gene expression microarray data obtained from post mortem prefrontal cortex of bipolar disorder and schizophrenia patients and matched controls from independent experiments. Comprehensive integration of microarray data has been carried out on the single probe level based on the sequence of each probe assayed on the microarrays. A number of computa- tional techniques have been utilized to carefully check the quality of the data and to infer and interpret the expression differences between the investigated conditions. Additionally, several tech- niques, including a novel Bayesian algorithm, have been used to cluster the patients according to their gene expression profile. We found 535 and 213 genes specifically altered in bipolar dis- order and schizophrenia, respectively. Of these, only 30 genes were shared between the syndromes. In addition, we identified genes of the transcriptional and post-transcriptional machineries altered in BD and genes of the development changed in SZ, suggesting that these functions can be distinctive biological markers of the two conditions. However, we failed to cluster the samples according to the diagnosis, probably due to the complex anatomical architec- ture of the prefrontal cortex and the resolution levels allowed by the microarray technology. Our findings suggest that, if SZ and BD share a common genetic pattern, the interaction with the environment and consequently the modulation of the gene expression can lead to different clin- ical features. The current classification of BD and SZ can be thus considered not only a useful model in the clinical practice, but it underlies real biological entities. doi:10.1016/j.nbt.2010.01.183 [P2.21] Elucidating the chromatin architecture of loci associ- ated with blood traits and coronary artery disease D.S. Paul 1,∗ , N. Soranzo 1,2 , W.H. Ouwehand 1,3 , P. Deloukas 1 1 Wellcome Trust Sanger Institute, UK 2 King’s College London, UK 3 University of Cambridge and National Health Service Blood and Transplant, UK Genome-wide association studies (GWAS) have identified com- mon genetic variants associated with common diseases, including coronary artery disease (CAD) and its main complication, myocardial infarction (MI). In parallel, GWAS on quantitative haematological traits, such as the count and volume of blood cells, have further elucidated through which biological process such disease-associated loci are likely to act. However, in most of these identified loci, which have usually modest effects, the causative variants remain unknown. Causal variants are very likely to impact regulation of gene expression. Nucleosome depletion is a common marker for active regulatory elements. Mapping accessible chromatin regions at loci associated with CAD/MI and relevant quantitative traits can thus aid identification of functional variants. We performed FAIRE (Formaldehyde-Assisted Isolation of Regulatory Elements), a robust method for identifying regulatory elements in human chromatin, on NimbleGen 385K tiling microarrays (9.6 Mb) in a megakaryocytic (MK) and erythrocytic cell line. In the MK cell line, we found 259 FAIRE peaks in 55 of 61 tested association loci. The peaks were prevalently located in inter- genic and intronic genomic regions (96%), and within 20 kb to the closest transcription start site (64%). Our experiments showed evidence of cell type specificity (P = 0.013), when comparing the number of FAIRE peaks in exclusively expressed genes in the megakaryocytic, erythroid, and monocytic lineages. At a genetic locus associated with mean platelet volume and function, we found megakaryocyte-specific FAIRE enrichment at the region of the lead SNP. This finding provides a first step in shedding light on a possible genetic mechanism. Investigating the sequence variation underlying nucleosome- depleted regions could guide the search for functionally relevant SNPs. In vitro assays could confirm the presence of regulatory ele- ments, and unravel its potentially altered function depending on genotype. doi:10.1016/j.nbt.2010.01.184 [P2.22] Prospective validation of a prognostic gene expression signature and identification of EGFR as a drug target in uveal melanoma V. Mirisola 1,∗ , P. Perri 1 , S. Coupland 2 , C. Mosci 1 , M. Truini 1 , U. Pfeffer 1 1 National Cancer Research Institute, Italy 2 University of Liverpool, UK Intraocular melanomas represent 5% of all melanomas and about 30% of patients develop metastases, usually in the liver, during the first five years after diagnosis. Uveal melanoma prognostic factors are histopathological and cytogenetic features. Onken et al. (Cancer Res 2004) showed that gene expression profiling identifies two dis- tinct molecular classes of uveal melanoma with low and high risk of metastasis. We describe here the prospective validation of the classifier and the identification of therapy targets through mRNA and miRNA expression profiling. Expression profiling was performed on 31 samples using Affymetrix HGUPlus2 arrays. Microarrays for miRNA screening were produced using the Exiqon library version 10.0. miRNA analyses were carried out on 28 samples. Differentially expressed miRNA and mRNAs were validated with qRT-PCR. Unsupervised hierarchical clustering of the samples using the classifier genes proposed by Onken et al. yielded three main clusters one of which contained all the six samples derived from tumors that developed metastases after a mean follow-up of two years. However, distance of the clusters was very limited indicating that the discrimination is not robust. A further up-date of the follow-up will be shown at the conference. mRNA expression analysis identified highly differential expres- sion of several drug targets among which EGFR, a target of specific kinase inhibitors and therapeutic antibodies. We found a signif- icant inverse correlation between the expression miR-128 and S64 www.elsevier.com/locate/nbt