Recent advances in genotyping and phenotyping technologies can now be applied in durum wheat and permit to adress a genome wide approach to disclose the geneti c architecture of kernel quality traits. DNA chips can allow the genotyping of thousands of SNP markers on several hundreds of lines while NIRS (Near Infra Red spectroscopy) allow the predicti on of relevant traits in durum wheat quality, e.g., protein content, yellow pigment content and semolina yield. Genome Wide Assocati on Studies (GWAS) have proven their potenti al in detecti ng chromosomal areas implied in the variati on of traits. Good panels should have a diversity level, a low structurati on and reduced level of linkage disequilibrium. Assembling panels from lines of diverse origin is usually the rule in many crops but may lead to a substanti al structurati on eff ect and to spurrious associati ons. We propose here to use lines drawn from an original composite cross populati on, originally dedicated to create a pre breeding resource. The Evoluti onary Prebreeding pOpulati on (EPO) is conti nuously grown since 1997 with a light selecti on and a 10% outcrossing rate. We report the fi rst results of a GWAS study on 300 hundred lines drawn from the 2009 harvest of EPO. Using a set of 62 000 polymorphic SNP drawn along the genome, we report the associati ons found between this set of markers and a set of traits related to the kernel quality predicted by NIRS. First, the characteristi cs of the EPO populati on will be illustrated to examine its power to detect associati on. We will report the fi ne geneti c structurati on and the fi ne scale Linkage disequilibrium on the 3B chromosome on which the SNP can be physically located. Then kernels quality traits will be dissected by GWAS to detect the most important chromosomal regions involved in the variati on. Finally, each trait will be predicted as the whole level to discuss the potenti al of Genomic Selecti on (GS) in the improvement of quality in the durum kernel.