How is molecular orbital theory used in drug research?

Question

I was reading up on MO theory and came across a particular section:

Application: Computational Chemistry In Drug Design

While the descriptions of bonding described in this chapter involve many theoretical concepts, they also have many practical, real-world applications. For example, drug design is an important field that uses our understanding of chemical bonding to develop pharmaceuticals. This interdisciplinary area of study uses biology (understanding diseases and how they operate) to identify specific targets, such as a binding site that is involved in a disease pathway. By modeling the structures of the binding site and potential drugs, computational chemists can predict which structures can fit together and how effectively they will bind.

Whilst I can see from the idea of in-phase and antiphase combinations of wave functions that you could potentially 'match' two atomic orbitals together but I am confused as to how this happens over a large molecule and the binding site of the target?

Answer 1

Common drug design approach does not involve much QC. There are two general approaches - QSAR and docking

QSAR

The idea is that a family of chemicals (training set) with some having desired biological activity, is used to train a pattern recognition algorithm. The trained algorithm is then used on a library of candidates and once recognised as active are then synthesised and tested in a lab.

Docking

The idea is to construct a molecule strongly interacting with a biological important target (a receptor or an enzyme). Since intermolecular interaction are weak, their calculation from first principles is very computationally expensive and is not practical in most cases. Since biologically relevant molecules are usually fairly trivial, even if complex molecules (For example, there is usually no need to consider interactions involving continuous conjugated systems), the intermolecular interactions are easily described using approach of molecular mechanics. So, in this approach a set of molecules is generated and then tested for interaction with a protein target. The tight spot is to find a way two molecules can bind, typically involving search over extremely large conformer space. Billions of potential candidates for ligand-protein complex must be tested, so the computationally cheapest solution (molecular mechanics) is used.

========

As you can see, neither approach normally uses orbital theory or any complex QC, and even if such complex QC is involved, it is clearly subservient. Instead machine learning algorithms and ways to efficiently search conformation space for low-energy conformers are the focus of research.

Answer 2

Given a crystal structure of a protein, some people are starting to do QM/MM modelling of the binding pocket. It's necessary to assume that the protein does not change conformation during binding, to keep the computation time reasonable. Unfortunately, this assumption is often false.