The constantly accelerating speed of microbial genome sequencing keep on ?ooding public databases with sequences of predicted proteins, only a small portion of which has been ever investigated experimentally or could be studied in detail any time soon. The only feasible way to annotate functions to these proteins is to predict them through bioinformatics analysis. Founded in 1997, the Clusters of Orthologous Groups of proteins (COGs) database has been a widely used tool for functional annotation. The COG database which are classified into 26 functional categories went through several updates, and gradually expanded its genome coverage to 62 organisms involved in 46 bacterial, 13 archaeal and 3 eukaryotic genomes.