Altering the characteristics of an active-site loop in an (S)-selective ω-transaminase from Arthrobacter citreus (variant CNB05-01) influences the enantioselectivity. This active-site loop belongs to the second subunit of the dimeric enzyme structure that participates in the coordination of pyridoxal-5′-phosphate (PLP) in the so called “phosphate group binding cup”. Three amino acid residues (E326, V328, and Y331) in this loop are selected by homology modeling for site-directed mutagenesis aiming to increase the enzyme enantioselectivity for 4-fluorophenylacetone. By combining these mutations, five enzyme variants are created. The performance of these variants is explored using a model system consisting of isopropylamine and 4-fluorophenylacetone or 4-nitroacetophenone in asymmetric synthesis using a whole-cell system approach. Three of the five variants show increased enantioselectivity for 4-fluorophenylacetone compared to CNB05-01. Variant CNB05-01/Y331C increases the enantioselectivity from 98 % ee to over 99.5 % ee. A single-point mutation, V328A, turn the (S)-selective ω-transaminase into an (R)-selective enzyme. This switch in enantioselectivity is substrate dependent, exhibiting (R) selectivity for 4-fluorophenylacetone and retaining (S) selectivity for 4-nitroacetophenone. The shift in enantiopreference is further confirmed by molecular docking simulations. Homology modeling is shown to be a powerful tool to target important amino acid residues in this enzyme in order to improve enantioselectivity by rational design.