AlgoGen: A tool coupling a linear-scaling quantum method with a genetic algorithm for exploring non-covalent interactions

摘要

AlgoGen-DivCon is a program that greatly benefits from algorithmic advances in quantum chemistry. It was initially designed to perform rigid molecular docking in order to ultimately pose a ligand in the receptor site by combining the Divide and Conquer linear-scaling quantum-chemistry method with a genetic algorithm (GA). A new version of this program with several enhancements is presented, interfaced with MOPAC/MOZYME. A biological application on seven docking structures leads to a pose in good agreement with known crystallographic structures. But, more generally, AlgoGen can explore intermolecular potential energy surfaces without preconceived idea, what yields an alternative use of this program. This feature was employed to investigate the possible presence of computational artefacts on the semiempiricial PM6-DH+ potential energy surfaces (PES) of 22 relative small complexes. For all dimers, the PM6-DH+ PES features a minimum geometry almost identical to the high-level reference equilibrium geometry. This method is found to perform remarkably well in predicting properties of hydrogen bonded complexes. However, in addition to the expected minima, false positive structures associated with wellcharacterized minima on the PES were identified for the ammonia and water dimers. Detection of these artefact makes AlgoGen PES scans an interesting tool for semi-empirical method developments aiming at reproducing non-covalent interactions and their evaluation. Additionally, a complementary post-treatment using NCI analysis turns out to give significant insight into chemical weak interactions found by AlgoGen.