Poster 44: Virtual Screening for Sirtuin Inhibitors using Target-specific Docking Methods.Michael Scharfe1, Martin Pippel1, Benjamin Maurer2, Manfred Jung2, Wolfgang Sippl1
|1Institut für Pharmazie, Martin-Luther-Universität Halle, Germany|
2Institut für Pharmazeutische Wissenschaften, Albert-Ludwigs-Universität Freiburg, Germany
|The Sirtuin (SIRT) family of human proteins are important regulators of many intracellular processes such as gene transcription and metabolism. They catalyze the cleavage of NAD+ and the subsequent transfer of certain functional groups from their target substrates to ADP-ribose. Sirtuin substrates are linked to the pathogenesis of diseases like cancer, neurodegeneration and metabolic disorders hence they are relevant drug-targets for small molecule inhibition.|
While for other NAD+-dependent enzymes a number of potent and selective inhibitors have been discovered, only a small number of mostly subtype-unspecific sirtuin inhibitors are available so far. We focussed our studies on the less well characterized subtypes SIRT5 and SIRT4 to improve the knowledge on their physiological function and to guide the identification of isoform-specific SIRT inhibitors.
SIRT4 and SIRT5 are located in the mitochondria and target predominantly metabolic enzymes like glutamate dehydrogenase (SIRT4) and carbamoylphosphate synthetase 1 (SIRT5). As recently reported, SIRT5 removes posttranslational succinyl and malonyl modifications of lysine residues by showing an unique specifity for negatively charged acyl groups. Using this knowledge and the available crystal structures of SIRT5, we have applied a combination of structure-based molecular modeling and in-vitro assay experiments for discovering SIRT5 specific inhibitors .
Active ligands from initial screenings and predictions concerning their putative binding mode were integrated into subsequent virtual screening cycles. Molecular interactions are investigated by means of docking in order to derive information about the underlying binding mechanism. General ligand interactions at the NAD+ binding site and putative binding modes of discovered inhibitors are incorporated into the development of target-specific scoring methods as a part of our open-source docking program ParaDockS .
Therefore, we collected publicly available crystal structures of NAD+ binding proteins and derived targeted pair potentials which can be used as scoring function within ParaDockS. Furthermore, target-specific interactions like the succinylated lysine recognition by SIRT5 can be included as pharmacophore type contraints to guide the conformational sampling of the docking algorithm. Such constraints were found to be helpful for proofing binding hypotheses and can be used as an optional filter for virtual screening.
As recently shown, protein-ligand interactions obtained from crystal structures as well as molecular docking poses can be used as input features for supervised learning algorithms (e.g. support vector machines or random forests). We have tested such classifiers on docking results of established datasets and could observe a classification performance superior to classical docking scores. Therefore, protein-ligand interactions obtained from molecular docking poses provide an adequate input for learning post-docking-classifiers. This outcome is an important aspect for using machine-learning algorithms on less well characterized proteins with only a few number of known ligands. In current studies we apply these newly developed docking and classification approaches in order to find selective inhibitors for human SIRT5 and SIRT4. Most of the methods are implemented in the ParaDockS framework which is distributed under the terms of GNU GPL2.
 B. Maurer, T. Rumpf, M. Scharfe et al.: Inhibitors of the NAD dependent protein desuccinylase and demalonylase Sirt5. ACS Med Chemistry Lett, 3 (12), 1050–1053, (2012)
 R. Meier, M. Pippel, W. Sippl et al: ParaDockS – A framework for Molecular Docking with Population-Based Metaheuristics. J. Chem. Inf. Model., 50:879-889, (2010)