The research carried out in my group centres around the use of ab initio and density functional theory (DFT) calculations, along with analysis of the resulting charge density distributions, to study and interpet a wide range of chemical phenomena. Much recent effort has gone into the study of intermolecular interactions, but we are interested in using similar methods to evaluate bonding and reactivity in a wide range of organic, inorganic, and biological systems.
In recent years, we have focussed on developing and testing methods to study non- covalent interactions such as hydrogen bonding and stacking, with particular interest in their importance for the structure of biological molecules and drug-receptor interactions. Ab initio methods such as MP2 and CCSD(T) give excellent accuracy for many such interactions, but in their conventional form cannot be applied to systems containing more than a handful of atoms. We have shown how use of local correlation methods such as local MP2 (LMP2) can reduce the computational cost of these methods without loss of accuracy for description of non-covalent interactions. The example below shows how this LMP2 approach accurately reproduces the potential energy surface for a stacked benzene dimer.
Since demonstrating the potential of this approach for model systems, we have applied it to a wide range of biologically important phenomena. The stacking of co-parallel bases in DNA and RNA is one such example: the efficiency of the LMP2 approach allowed us to evaluate benchmark ab initio interaction energies for different base-pair steps, against which more approximate methods could be tested. We have also used the same methods to predict the binding of intercalators, an important class of anti-cancer drug, to DNA, highlighting the structural and energetic properties that affect binding. A typical intercalator, ethidium bromide, bound to a base pair step, is shown below.
As well as accurate ab initio methods, we are interested in applying faster, more approximate methods to study of non-covalent interactions, in order to further extend the size of molecules or cluster that may be treated. Conventional density functional theory (DFT) performs poorly for many non-covalent interactions, for example failing to predict any binding whatsoever between stacked systems. In recent years we have shown how judicious choice of the parameters of hybrid DFT can give surprisingly accurate predictions of the geometry and energy of molecules and clusters bound by non-covalent interactions. In particular, Becke's half-and-half functional reproduces interaction energies and potential energy surfaces for a wide range of stacked molecules. Much of this work has concentrated on DNA and its interactions with metal based anti-cancer drugs such as cisplatin. These bind covalently to nucleophilic sites within the DNA chain, with complementary non-covalent interactions in many cases. This binding disrupts the classical double-helix structure of DNA to such an extent that repair proteins induce programmed cell death, or apoptosis. A suitable theoretical description of this drug-receptor interaction must therefore properly describe both the metal-DNA covalent and non-covalent bonding, and the internal non-covalent interactions within DNA itself. An example of cisplatin binding to two guanine bases is shown below.
Although DFT is rather computationally efficient, the size of representative portions of DNA means that one cannot hope to model an entire double helix, along with associated counterions and solvent, interacting with such a drug. In this case, we turn to hybrid quantum mechanics/ molecular mechanics (QM/MM) methods, in which only a small part of the entire system is treated with a quantum mechanical method such as DFT, with the remainder approximated by a much faster, classical approach. Appropriate classical approaches such as AMBER and CHARMM are well-validated for nucleic acids, and allow efficient simulation of much larger DNA fragments than would be possible otherwise. A typical set-up of a QM/MM calculation for cisplatin interacting with DNA is shown below, with the QM region in bold and MM region wireframe.
Platinum is not the only metal of interest in this field; recent work by the group of Peter Sadler has shown that ruthenium-arene complexes show a great deal of promise as potential anti-cancer agents. They also bind to DNA through a combination of covalent and non-covalent interactions, but here the non-covalent part is rationally designed to induce greater disruption through stacking interactions with DNA bases. DFT optimised geometry of one such adduct, showing the co-planar orientation of arene ligand with DNA base, is shown below.
As well as these structural and energetics studies, I maintain an interest in using quantum mechanical calculations coupled with statistical methods to predict solvation properties. For instance, we have measured the permeability of around 40 organic and drug compounds through membranes consisting of 2% dioleyl- phosphatidylcholine (DOPC) in dodecane, which is widely used as a screen for in vivo cell and intestinal permeability. Using the inherent pH dependence of this permeability (see below, left), we were able to calculate the "inherent" permeability of each compound, which was then modelled using the established statistical methods. As a result, we now not only have a predictive model suitable for virtual screening of new compounds, but also physical insight into the molecular properties that inhibit or promote membrane permeation (below, right).
In addition to the rather applied problems described above, I have long-standing interests in studying a variety of chemical bonding and reactivity problems using theoretical methods. In this, we make extensive use of Bader's theory of Quantum Theory of Atoms in Molecules (QTAIM). Much of this work is carried out in collaboration with crystallographers such as Assoc. Prof. David Hibbs (Sydney) and Dr. Jacob Overgaard (Aarhus), who are able to measure the same properties that our calculations predict. Systems of interest vary from biologically relevant molecules such as the neurotransmitter taurine to novel inorganic species such as subvalent,low-coordinate group 13 and 15 compounds. The latter are carried out as part of a collaboration with Prof. Cameron Jones (Monash). Recent highlights include the first characterisation of a non-nuclear maximum in the electron density in a molecule, namely a Mg(I)-Mg(I) dimer. This highly unusual feature has been predicted for body-centred cubic alkali and alkaline earth metals, but its presence in a stable molecule is unprecedented and remarkable. The non-nuclear maximum is observed in both experimental and calculated electron density, and is not an artefact of DFT or multipole methods employed to construct these densities. The presence of this maximum also gives rise to a large area of charge concentration, or negative Laplacian of the density, within the Mg-Mg bond, as shown below.
In another, we compared a number of density-based properties for the important neurotransmitter taurine, a zwitterionic compound containing a sulphonic acid group. It was found that hydrogen bonding interactions in the crystal significantly deform the bonding in taurine compared to the free gas-phase molecule, results which may be important for the interactions between taurine and lipids, receptors, and other biomolecules in the condensed phase. The electrostatic potential around taurine is shown below.