T dataset. The chemical structures of these 7385 compounds, for which a target protein was identified in the PDpB, were downloaded as ideal CCD (Chemical Compound Dictionary) coordinates (http:www.wwpdb.NFPS manufacturer orgccd.html).Compound PromiscuityCompounds bound to 3 or a lot more non-redundant target pockets have been defined “promiscuous,” all other folks “selective.”DrugsChemical structures of all non-nutraceutical little molecule drugs (authorized and experimental) were downloaded as structure-data files (SDF) in the DrugBank database (Wishart et al., 2006) (D-4-Hydroxyphenylglycine Autophagy version 4.1, 20140908) comprising a total of 6858 drug molecules.Compound Classification and Home CalculationMolecular weights and SMILES strings (“Simplified Molecular Input Line Entry Specification”) of all compound structures were calculated utilizing the Instant JChem computer software (version 14.7.7.0, ChemAxon, http:www.chemaxon.com). Really compact or huge compounds (molecular weight 30 Da or 1000 Da), variable compound structures comprising R-groups and compounds devoid of computable SMILES had been not consideredProtein Targets and Co-crystallized CompoundsTo produce the protein target set linked with all compounds, all available protein structures with at the very least one particular co-crystallized, non-covalently bound compound plus a X-ray crystallographicFrontiers in Molecular Biosciences | www.frontiersin.orgSeptember 2015 | Volume two | ArticleKorkuc and WaltherCompound-protein interactionsfor further analysis. The chemical improvement kit (CDK) extended fingerprints in the rcdk R-package (Guha, 2007) was employed for similarity evaluation of compound structures. Drugs or metabolites have been mapped to PDB compounds requiring identical molecular weights (at 1 Da tolerance) and identical fingerprint (Tanimoto distance, T, T 0.95; 91 of all compounds mapped with T = 1.0). PDB compounds assigned to both drug and metabolite compounds have been labeled as “overlapping compounds.” Physicochemical properties of those the compound class regarded right here (drugs, metabolites, and overlapping compounds) had been calculated by utilizing Instant JChem and KNIME (Berthold et al., 2008) (version 2.9.4) (The list of all computed properties is offered in Supplementary Figure 1). Properties based on actual 3D-structures were depending on the best Chemical Compound Dictionary (CCD) compound coordinates (http:www.wwpdb.orgccd.html).errors, se, on the obtained propensities had been calculated as defined in Levitt (1978) with: sei = Propensity values have been symmetrical distributions. 1 gi fi (1 – fi ) n i = 1 qi log10 -transformed to (3) produceAmino Acid Residue Compositional Propensities of Protein Binding PocketsCompound binding pocket amino acid composition propensities were calculated working with Equation (2), followed by log10 transformation and with qi representing the number of amino acid residues of form i = 1, …, 20 in binding pockets and si the number of amino acid residues i = 1, …, n in non-binding internet site components of proteins.Compound-promiscuity Propensity Ratio CalculationPhysicochemical properties preferentially connected with either promiscuous or selective compounds (Table 1B) had been judged depending on propensity values, P, calculated for each property variety t and compound class c as: Pit,c = qi fi = gi si n i = 1 qi , n i = 1 siEnzyme Classification Entropy and Pocket Variability AnalysisThe degree of target set variability linked with each promiscuous compound was characterized by two measures, the entropy of EC numbers of target proteins along with the variabili.