Fig. 1

Development of the DAIRYdb consisted in three main steps: construction, curation and validation. For construction, dairy products specific 16S sequences were retrieved from Silva, RDP and Greengenes using Genbank NCBI, EMBL, Agroscope and INRA sequences. Curation was performed based on the cross-validation results from the leave-one-out test of SATIVA and highly iterated RAxML tree, followed by manual curation of taxonomic assignment and consistency throughout all taxonomic ranks, with a particular focus on singleton taxons with no reference sequence. Validation was performed comparing taxonomy annotation accuracy of single and HVR pairs by the five databases (Greengenes 13.8, LTP version, Silva 128 NR99, RDP version and DAIRYdb)