The microbiota in the rumen fluid of 65 Holstein cows was characterized through the 16S rRNA (bacteria) targeted re-sequencing. Methane yields (CH4y = CH4/DMI) and dry matter intake (DMI) were individually measured during 3 weeks with the GreenFeed and Calan gate systems, respectively. Twenty of these cows were measured (CH4y) and rumen contents sampled twice in two successive months. A combination of multivariate, clustering and microbial co-abundance network analysis was implemented to identify a set of OTUs jointly associated with CH4y and the discrimination of samples according to the structure of ruminal microbial ecosystem. In addition, repeated measures were used to provide an estimation of the 16S ruminal stability within a month, which was R=0.64. Three ruminotype-like clusters (R1, R2 and R3) were identified; and R2 was associated with higher CH4y emission. Sparse partial least squares discriminant analysis (sPLS-DA) was done based on samples classification following two criteria: ruminotype-like cluster assignation and CH4y phenotype classification (low, high and intermediate). A sPLS regression model using the phenotype variation of CH4y emission was also implemented. Overall, 28 OTUs, which explain the 52% of the phenotype variation of CH4y, were identified by combining the approaches. The taxonomic classification of these OTUs included families linked to CH4 emission such as Christensenellaceae, Lachnospiraceae, and Ruminococcaceae. According to the interaction patterns, 83% (25/28) of these OTUs belong to the same sub-network module. In summary, our results suggest a relative stability of the rumen microbiota as well as a common set of interacting OTUs simultaneously linked to CH4y emission and microbial ruminal ecosystem structure.