The size of a typical P450 enzyme (approximately 3500 atoms) is
prohibitively large to be modelled in its entirety using an ab initio
approach at present, even with the efficient methods described in Chapter
running on modern supercomputers. Therefore, attention
must be focused on a smaller region of the enzyme around the active
site. Indeed, an important aim of this investigation will be to
determine if many properties of an enzyme may be modelled by
consideration of only a small portion of the entire molecule. It was
desirable to limit the computational resources required and to address
the properties of P450s in general, therefore none of the contact
residues which define the substrate binding pocket were included in
the simulations. The model systems contained only the ligand molecule,
haem moiety, 
-ligated cysteine residue and -ligated water (if
present). Examples of substrate-free, substrate-bound,
inhibitor-bound, and substrate analogue-bound systems modelled are
shown in Figures , , and
respectively. No questions of
substrate specificity were addressed as this is defined by the
geometry of the substrate binding pocket. In the future, questions
regarding the nature of the 
-bound system and also specific
P450 enzyme species may be addressed. This will require the inclusion
of additional contact residues.
Due to the limited size of the system modelled and the uncertainty in
the structure of the active sites of most P450s, cytochrome 
(CYP101) was chosen as the subject of this investigation
because accurate crystal structures of this enzyme have been
obtained. These structures include the enzyme complexed with
substrates, substrate analogues and inhibitors. The systems included
in this study are listed in Table . The use of
crystal structures allows the determination of the position of the
ligand relative to the active site and the changes in the geometry of
the active site caused by ligand binding. This restricts one source of
uncertainty in the results obtained from the calculations. At this
early stage it is important to limit any possible external sources of
error so that any deviations from experimental data can be attributed
to the computational modeling approach taken in the simulation.
| Ligand | Nature of Ligand | PDBreference | Reference |
| Substrate-free | N/A | 1PHC | [93] |
| Camphor | Substrate | 4CP4 | [100] |
| Adamantanone | Substrate | 5CPP | [101] |
| Camphane | Substrate analogue | 6CPP | [100] |
| Norcamphor | Substrate analogue | 7CPP | [101] |
| Metyrapone | Inhibitor | 1PHG | [98] |
This investigation addresses two questions. Firstly, it is important
to determine if our ab initio methods accurately reproduce the
changes in the spin state of the observed experimentally. This
serves to validate the use of our approach for the study of the P450
system and will enable more speculative questions to be addressed with
confidence. In particular, the rôle of the -coordinated water
molecule, found in the substrate-free and substrate analogue-bound
complexes, is explored.
The calculations were performed with a 600eV cut off energy for the plane wave basis set using a spin-dependent GGA [18] to the exchange-correlation potential. The systems were enclosed in a cubic supercell with a side of length 19Å and contained between 90 and 118 atoms. A non-linear core-corrected pseudopotential was used to describe the ionic core of the iron atom. The main spin dependent calculations were performed on 64 processors of the Hitachi SR2201 parallel supercomputer at the Cambridge High Performance Computing Facility (HPCF). Preliminary calculations were performed on 64 nodes of the Hitachi SR2001 supercomputer at the Maidenhead headquarters of Hitachi Europe Ltd and on 16 nodes of the SR2201 at the Cambridge HPCF. These preliminary calculations primarily involved the relaxation of the hydrogen atoms in the systems into their equilibrium positions as their atomic coordinates were not specified in the crystal structure datasets.