Simple filter Simple molecular properties (e.g., mole- 103 to Is cular weight, molecular volume, element composition, etc.)
Complex filter Molecular properties that are more com- 0.1 to 10 s plicated to compute (basicity, acidity, drug-likeness, etc.)
Atom types and substructures in a binary representation Lipophilicity, pK,,-values
Physico-chemical properties Substructure descriptors
Autocorrelation coefficients Quantum chemical descriptors Feature - feature distances
Three-dimensional pharmacophore patterns
Structure - activity relationship-
Patterns of substructures (functional groups, atom types, chains, ring systems, etc.)
Topology, branching, general shape 0.1 to 10 s
General shape, distribution of atomic 1 to 100 s properties
Charge distribution in the molecules, 10 to 102 s molecular interaction fields
Representation of certain pharmaco- 1 to 102 s phore patterns
Target interaction sites (hydrogen bonds, 10 to 102 s electrostatic or lipophilic interactions, etc.). Similarities between different chemical classes and scaffolds can be recognized. Pharmacological activity 10tol02s
Similarity of shape and similarity of pos- 102 to 103 s sible interaction sites. Similarities between different chemical classes. Pharmacological activity 102 to 103 s
6. Descriptors based on the three-dimensional structure of molecules are applied when sets of structurally diverse compounds must be compared. The time needed for the calculations may exceed several minutes or even hours.
7. If the three-dimensional structure of the biological target (an enzyme or receptor that is to be affected by the hypothetical drug molecule) is known, it might be informative to test every structure of the library for a steric fit to this target. Such calculations take at least a few minutes for each structure and are therefore only feasible for relatively small virtual libraries.
Was this article helpful?