Pyridoxal 5'-phosphate (PLP) dependent enzymes comprise a large family that plays key roles in amino acid metabolism and are acquiring an increasing interest as drug targets. For the identification of compounds inhibiting PLP-dependent enzymes, a chemogenomics-based approach has been adopted in this work. Chemogenomics exploits the information coded in sequences and three-dimensional structures to define pharmacophore models. The analysis was carried out on a dataset of 65 high-resolution PLP-dependent enzyme structures, including representative members of four-fold types. Evolutionarily conserved residues relevant to coenzyme or substrate binding were identified on the basis of sequence-structure comparisons. A dataset was obtained containing the information on conserved residues at substrate and coenzyme binding site for each representative PLP-dependent enzyme. By linking coenzyme and substrate pharmacophores, bifunctional pharmacophores were generated that will constitute the basis for future development of small inhibitors targeting specific PLP-dependent enzymes.