Since its release in 199010 (link), AutoDock has proven to be an effective tool capable of quickly and accurately predicting bound conformations and binding energies of ligands with macromolecular targets9 ,11 (link)–14 . In order to allow searching of the large conformational space available to a ligand around a protein, AutoDock uses a grid-based method to allow rapid evaluation of the binding energy of trial conformations. In this method, the target protein is embedded in a grid. Then, a probe atom is sequentially placed at each grid point, the interaction energy between the probe and the target is computed, and the value is stored in the grid. This grid of energies may then be used as a lookup table during the docking simulation.
The primary method for conformational searching is a Lamarckian genetic algorithm, described fully in Morris et al.9 . A population of trial conformations is created, and then in successive generations these individuals mutate, exchange conformational parameters, and compete in a manner analogous to biological evolution, ultimately selecting individuals with lowest binding energy. The “Lamarckian” aspect is an added feature that allows individual conformations to search their local conformational space, finding local minima, and then pass this information to later generations. A simulated annealing search method and a traditional genetic algorithm search are also available in AutoDock4.
AutoDock4 uses a semiempirical free energy force field to predict binding free energies of small molecules to macromolecular targets. Development and testing of the force field has been described elsewhere11 (link). The force field is based on a comprehensive thermodynamic model that allows incorporation of intramolecular energies into the predicted free energy of binding. This is performed by evaluating energies for both the bound and unbound states. It also incorporates a new charge-based desolvation method that uses a typical set of atom types and charges. The method has been calibrated on a set of 188 diverse protein-ligand complexes of known structure and binding energy, showing a standard error of about 2–3 kcal/mol in prediction of binding free energy in cross-validation studies.
The primary method for conformational searching is a Lamarckian genetic algorithm, described fully in Morris et al.9 . A population of trial conformations is created, and then in successive generations these individuals mutate, exchange conformational parameters, and compete in a manner analogous to biological evolution, ultimately selecting individuals with lowest binding energy. The “Lamarckian” aspect is an added feature that allows individual conformations to search their local conformational space, finding local minima, and then pass this information to later generations. A simulated annealing search method and a traditional genetic algorithm search are also available in AutoDock4.
AutoDock4 uses a semiempirical free energy force field to predict binding free energies of small molecules to macromolecular targets. Development and testing of the force field has been described elsewhere11 (link). The force field is based on a comprehensive thermodynamic model that allows incorporation of intramolecular energies into the predicted free energy of binding. This is performed by evaluating energies for both the bound and unbound states. It also incorporates a new charge-based desolvation method that uses a typical set of atom types and charges. The method has been calibrated on a set of 188 diverse protein-ligand complexes of known structure and binding energy, showing a standard error of about 2–3 kcal/mol in prediction of binding free energy in cross-validation studies.
