The authors present a method for assigning protein functions based on comparative genome analysis. They propose that proteins that function together in a pathway or structural complex evolve in a correlated manner, leading to the preservation or elimination of these proteins in new species. This property is characterized by creating phylogenetic profiles for each protein, which are strings indicating the presence or absence of a protein in every known genome. Proteins with matching or similar profiles are likely to be functionally linked. The method is applied to the genome of *Escherichia coli* to predict the functions of uncharacterized proteins. The results show that proteins with similar phylogenetic profiles are often functionally linked, and the method can accurately predict the functions of uncharacterized proteins by examining the functions of their phylogenetic-profile neighbors. As more genomes are sequenced, the length and informativeness of these profiles will increase, making the method more powerful for studying protein functions in various organisms.The authors present a method for assigning protein functions based on comparative genome analysis. They propose that proteins that function together in a pathway or structural complex evolve in a correlated manner, leading to the preservation or elimination of these proteins in new species. This property is characterized by creating phylogenetic profiles for each protein, which are strings indicating the presence or absence of a protein in every known genome. Proteins with matching or similar profiles are likely to be functionally linked. The method is applied to the genome of *Escherichia coli* to predict the functions of uncharacterized proteins. The results show that proteins with similar phylogenetic profiles are often functionally linked, and the method can accurately predict the functions of uncharacterized proteins by examining the functions of their phylogenetic-profile neighbors. As more genomes are sequenced, the length and informativeness of these profiles will increase, making the method more powerful for studying protein functions in various organisms.