Animal breeding has shown in recent years an increasing interest in the use of digestive microbiota information, given the association with the physiological traits of the host. In this context, the hypotheses of this thesis are: 1) the rumen microbiota influence milk composition and udder health traits, and 2) a genetic determinism exist on rumen microbiota that affects milk composition and udder health traits. To address these hypotheses, we explored the relationship between the host genome and ruminal metagenome in the modeling of dairy ewe’s phenotypes. The “MicroGenOL” project was based on 795 Lacaune ewes from two genetic lines reared under the same housing and feeding conditions at the INRAE Experimental Unit of La Fage. For all animals the fine milk composition was predicted by mid-infrared spectra, and for 118 ewes, fatty acids (FAs) were measured by gas chromatography. In addition, all animals were genotyping with 54k SNP chip, and one rumen sampling for all animals, or two one week apart for 118 ewes, were collected. Rumen samples were sequencing of 16S RNA gene. To work with microbial abundances due to their compositional nature, a workflow was defined: after bioinformatics process of the raw sequences, operational taxonomic unit (OTU) counts in the abundance table were transformed in logratios (after zeros imputation) and corrected by environmental factors using linear models. The variability of microbial abundances between two weeks and the impact of data processing on genetic estimations allowed for us to defined the best way to work with, and the limits of results interpretation. The first analysis focused on the influence of the rumen microbiota on the phenotype. We found low (