7

In the problem of protein-ligand docking, most of the time people are happy if they can just predict the final conformation the ligand adopts into the protein's binding pocket. Most of the time one can just use a physical based scoring function and an energy minimization method to try to find the minima. But in some other cases, scoring functions are very inaccurate to represent some more complex interactions; metal interactions, cation-pi, etc., and even covalent bonding. So it seems that only ab-initio methods can shed light here.

Can quantum mechanical methods be used in docking for this purpose or not and which are the best methods, even if you require access to HPC resources?

I am mostly interested in covalent bond formation. It would be great also if you could provide some references, I did not find many in the literature.

Peter Mortensen
  • 158
  • 8
Open the way
  • 711
  • 1
  • 9
  • 15

2 Answers2

4

It's been known for quite some time that quantum methods have a useful role to play in studying protein-ligand docking. However, the challenge that you have—which has not really changed too much in recent years—is that ab initio methods do not cope well with very large systems. It's probably possible to study the interaction of a ligand with the relevant binding sites of the protein, but when you're stuck at modeling a few hundred to a few thousand atoms on very large machines, it's difficult to study the entire interacting system (which would include not only the protein and the ligand, but quite possibly also the aqueous solvent as well!).

However, there are some references available: Gresh et al. have proposed a "bottom-up" molecular mechanics strategy that incorporates ab initio data, while Morelli et al. combine ab initio docking calculations with experimental analysis to move towards studying protein-protein complexes.

aeismail
  • 3,523
  • 1
  • 18
  • 31
2

There exist hybrid methods such as ONIOM and QUILD (a rough cognate of ONIOM) that address the interface between QM and MM calculations and are well-suited to investigation of protein cofactors and probably refining ligand docking models. The ONIOM methodology is very simple but clever - you break the calculation into the bulk protein which is handled with MM, and the region of interest, which is dealt with using an electronic structure method. Forces on capping atoms shared by both regions are reconciled between the QM and MM calculations, and relative energies are obtained algebraically.

This approach would not be suited to high-throughput docking studies and would instead be more useful for checking the viability of a few good leads. I know a guy who has built his PhD around using QUILD to (AFAIK) assess models of water binding modes to a protein active site. The space of possible solutions is reduced because the water has to datively bind to cofactor metals.

Richard Terrett
  • 233
  • 1
  • 6