Abstract

The use of computational quantum chemical methods to aid drug discovery is surveyed. An overview of the various computational models spanning ab initio, density function theory, semiempirical molecular orbital (MO), and hybrid quantum mechanical (QM)/molecular mechanical (MM) methods is given and their strengths and weaknesses are highlighted, focussing on the challenge of obtaining the accuracy essential for them to make a meaningful contribution to drug discovery. Particular attention is given to hybrid QM/MM and semiempirical MO methods which have the potential to yield the necessary accurate predictions of macromolecular structure and reactivity. These methods are shown to have advanced the study of many aspects of substrate–ligand interactions relevant to drug discovery. Thus, the successful parametrization of semiempirical MO methods and QM/MM methods can be used to model noncovalent substrate-protein interactions, and to lead to improved scoring functions. QM/MM methods can be used in crystal structure refinement and are particularly valuable for modelling covalent protein-ligand interactions and can thus aid the design of transition state analogues. An extensive collection of examples from the areas of metalloenzyme structure, enzyme inhibition, and ligand binding affinities and scoring functions are used to illustrate the power of these techniques.

Keywords: Quantum chemistry, QM/MM method, semiempirical MO method, carbohydrate, metalloenzyme, enzyme inhibitor, docking of substrate, dispersion correction.