Combining a genetic algorithm with a linear scaling semiempirical method for protein-ligand docking

In the search for more energetically precise docking methods and to avoid the problem of defining suitable ligand force fields, we have built a software to perform rigid molecular docking by combining a genetic algorithm with a linear scaling semiempirical program. The resulting AlgoGen-DivCon software uses the Divide and Conquer linear scaling methodology to evaluate protein-ligand interaction energies using quantum semiempirical hamiltonians while conformational search is performed by a genetic algorithm which optimizes the position of a mobile ligand relative to a fixed protein system. Tests have been performed to assess the performance of AlgoGen-DivCon. Our results on the docking of 8-azaxanthine, oxonic acid and uric acid in a 700 atom Urate Oxidase model are structurally very similar to known crystallographic structures and prove that it is now possible to perform in a reasonable amount of time molecular dockings of relatively large protein-ligand complexes using a quantum description of the interaction energy.