Résumé
A large sample (430 out of about 1650 genotypes) of rice genetic resources collected from European breeders has been described with 50 phenotypic (agronomical, morphological and technological) traits and 16 genotypic (micro-satellites or SSR) markers. On the basis of available passport information and 7 classic characteristics of plants (leaf width, form of the panicles, etc) these genotypes were classified a priori within 10 agro-morphological types. Two multivariate analysis (principal components) based on the one hand on the 7 morphological traits, and on the other hand on grain technological characteristics clearly identify three main groups : indica, "traditional" japonica and "North American" japonica. The distribution of some characters of interest and the correlations between some of them, reveal the strong impact of human selection. A first diversity analysis of the 430 genetic resources was performed using the Neighborjoining method of Saitou and Nei, on the basis of the sole SSR makers, leading to the construction of a diversity tree, which very clearly separates indica types and basmati types in opposition to all japonica types. The japonica types are themselves divided between around 100 homogeneous sub-groups. Four different methods were investigated in order to create a core collection of European rice genetic resources. The first consists in collecting all the existing SSR alleles in the population with the smallest number of entries (43 accessions. The second also contains all the existing SSR alleles but in selecting first the individuals with rare (frequency < 1%) alleles (84 accessions). The third approach considers the 100 homogeneous sub-groups which can be defined in the SSR diversity tree, and one entry is randomly selected in each sub-group (100 accessions). The last approach is based on a tree combining the genetic diversity of SSR alleles which provides the skeleton of the tree and the phenotypic diversity of agromorphological characteristics traits which is revealed in the more external part of the tree. As in the previous approach, one entry is randomly selected in each homogeneous sub-groups (117 accessions). The interest and the representativeness of these core collections are discussed, by comparing the distributions of original population which each one of the core collections for traits non used in the building process of core collections. (Résumé d'auteur)