Comparative genomics have shown that protein families vary significantly within and across organisms both in number and in functional composition. In the present work, we test whether the diversity at the family level reflects biological differences among organisms and contributes to their unique characteristics. For this purpose, we have collected sequence-similar proteins of three selected families from the model bacteria, Escherichia coli, Bacillus subtilis, and Pseudomonas aeruginosa. Protein relationships were identified using a phylogenomic approach in order to connect functional diversity of enzymes to the metabolic capabilities of these organisms. All protein families studied have distinct functional compositions across the selected bacteria as supported by our Bayesian analysis. Some conserved functional features among family members include shared reaction mechanism, cofactor usage, and/or ligand specificity. Many observations of presence/absence of protein functions match current knowledge of the physiology and biochemistry of the bacteria. In some cases new functional predictions were made to family members previously uncharacterized. We believe that genome comparisons at the protein family level would be useful in predicting metabolic diversity also for organisms that are relatively unknown or currently uncultured in the laboratory.