Recently, cocktail of bacteria were proposed in order to treat Clostridium difficile infection (CDI), but these bacteriotherapies were selected more by chance than experimentation. We propose to comprehensively explore the gut microbiota of patients with CDI compared to healthy donors in order to propose a consortium of bacteria for treating C. difficile. We compared stool samples composition from 11 CDI patients and 8 healthy donors using two techniques: metagenomics, 16S V3-V4 region amplification and sequencing and culturomics, high throughout culture using six culture conditions and MALDI-TOF identification. By culturomics, we detected 170 different species in the CDI group and 275 in the control group. Bacteroidetes were significantly underrepresented in the CDI group (p = 0.007). By metagenomics, 452 different operational taxonomic units assigned to the species level were detected in the CDI group compared to 522 in the control group. By these two techniques, we selected 37 bacteria only found in control group in more than 75% of the samples and/or with high relative abundance, 10 of which have already been tested in published bacteriotherapies against CDI, and 3 of which (Bifidobacterium adolescentis, Bifidobacterium longum and Bacteroides ovatus) have been detected by these two techniques. This controlled number of bacteria could be administrated orally in a non-invasive way in order to treat CDI.