To derive a compound set for virtual screening, an in-house virtual library containing 4,177,660 commercially available compounds was filtered by the following selection criteria: at least one but not more than five hydrogen-bond donors, at least one but not more than ten hydrogen-bond acceptors, at least nine but not more than 23 heavy atoms and a clogP between 21 and 4. In addition, the number of rotatable bonds was restricted to less than seven, the total charge between 21 and +1, and at least one but not more than two ring systems were allowed. Compounds containing unwanted functionalities were excluded. Only compounds that fulfilled all requirements were taken to the next step. The selected subset was further filtered using a protein-based pharmacophore. When deriving the pharmacophore we aimed to strike a reasonable balance between a complex query which potentially retrieves very potent compounds but has only a very low hit rate and a relaxed query retrieving many compounds which prove not be active. To not be over descriptive we decided to only include interactions to His25 which is essential for recognition of the cytosine moiety of the substrate. In all structures containing ligands interacting with this residue, ND presumably carries a hydrogen atom to hydrogen bond with the cytidine moiety of the ligands. However, in the crystal structures His25 NE is Staurosporine PKC inhibitor solvent exposed and not involved in a hydrogen-bonding network. Accordingly, it is possible that not ND but NE carries a hydrogen atom when challenged with ligands presenting a hydrogen-bond donor functionality. Therefore, both states were considered in the derived pharmacophore. Furthermore, hydrogen-bond acceptor interactions to the backbone amino group and hydrogen-bond donor interactions to the carbonyl group of His25 were required. 42,912 compounds fulfilled at least two of these pharmacophore features. In the final step, the remaining compounds were docked into the receptor and promising hits selected for testing. For this Nutlin-3 company purpose four different versions of the binding pocket were prepared taking into account different tautomers of His25 and the presence and absence of the co-factor. From each docking run, the top 200 scoring molecules together with the top 200 scoring molecules obtained when the score was divided by the number of heavy atoms were stored in the final hit list. The latter was done to favour small molecules which bind with a predicted high ligand efficiency. The resulting 1,600 docking poses were filtered for compounds still in agreement with the described pharmacophore hypothesis. Only 566 compounds fulfilled at least two interactions required by the pharmacophore. By visual inspection compounds with additional hydrogen-bonding or hydrophobic interactions to the binding site were favoured and finally 14 compounds were purchased for evaluation. Five of these were predicted to have interactions with His25 similar to those observed for cytidine while nine compounds satisfied the alternative arrangement of functional groups. In the modelled binding mode of ligand 4 the pyrrolopyridine scaffold is sandwiched between Tyr24 and Tyr175 and potentially interacts with His25 via three hydrogen bonds. In addition, the charged amino group in the tetrahydropyrimidinium ring may form a salt bridge interaction with Asp130. Commercially available analogues of this compound were selected to probe possible interactions with Asp130, His25 and the backbone of Lys145. Compounds 12 and 13 were chosen because of substituents on the tetrahydropyrimidinium ring which are likely to lower the pKa resulting in neutral compounds under assay conditions. Accordingly, these compounds can no longer form a salt bridge with Asp130.