Fflux for Protein Crystallography

The ab initio prediction of protein structure remains a holy grail of structural biology. One is even rarely able to accurately model or predict established structures or small variations thereof. The key issues here are the lack of reliable force fields, and the fact that protein and enzyme engineering studies or ligand docking studies require significantly higher accuracy of the model in order for it to be valuable (as compared to fold assignment or protein: protein docking modeling studies). Recent advances in computing and the generation of high-resolution crystal structures have made rigorous validation of a novel force fields against high-resolution (< 1.0 Å) crystal structures a distinct possibility. This is vital to make future-proof progress in the field of biomolecular mechanism and catalysis.

The shape of current biomolecular force fields (e.g. AMBER, CHARMM) was decided in the 1980s! We have taken advantage of the increase in computing power by four orders of magnitude since that time, in order to re-design force fields fundamentally. One of the developers of AMBER, wrote1 that “Without further research into the accuracy of force-field potentials, future macromolecular modeling may well be limited more by validity of the energy functions, than by technical ability to perform the computations.” In other words, without research such as proposed here, computers will, in future, deliver wrong predictions faster.

We developed novel force field (QCTFF) methodology2, now called FFLUX. There is the exciting prospect of validating FFLUX on high-resolution (crystal) structures. The influence of small changes, induced by mutations, determines if FFLUX predictions are trustworthy, and more importantly, right for the right reasons.

FFLUX abandons inherently limited point charges in favour of atomic multipole moments, ensuring the electrostatic part is tackled rigorously. The main idea is to construct “knowledgeable” atoms, drawn from small molecules and made to interact in order to predict properties of large molecules. These (topological) atoms3, 4 are 3D fragments of electron density that possess a finite volume. They have sharp boundaries, which endows them with an almost “malleable” character. The figure shows atoms in the capped oligopeptide GlyAlaThr in crambin. The atomic shape and properties respond to changes in the coordinates of the immediate environment of the molecule they are part of. The machine learning method Kriging can capture these changes.

In the MIB there are several projects where data to <1.0 Å are collected for enzyme complexes or mutants (PETNR reductase5 and P450 CYP1216). Also, there are more than 2000 structures with a resolution below 1.25 Å in the PDB. A key issue is the correct modelling of crystal structures, which are not accurately represented by the “model in a water box” approach often adopted in molecular dynamics. We will initially focus on parts of crystal structures, focusing on elements not involved in lattice contacts. The remainder of the structure will be kept fixed. This approach has the benefit of allowing an incremental increase in the size of the modelled structure and the possibility to select particular structural elements or select for particular amino acids. Correlation between modelled structure and the corresponding experimental coordinates is straightforward to establish and can be used to further develop FFLUX. Reliably modelling the effect of a point mutation on (local) protein structure or docking of ligands into active sites are obvious applications.

Qualifications
Applicants should have or expect a good II(i) honours degree (or an equivalent degree) in a Masters degree in Chemistry or Physics. The project is purely computational without tapping into mathematical skills or programming. If you have those skills then that is a bonus but it is not necessary. If you wish to develop these skills then this can be taken care of. However, the project requires the efficient and correct running of existing computer programs and a systematic analysis of the data generated.

Contact for further Information
For general admission information, please email: [Email Address Removed].

Informal inquiries about the project should be sent to Prof. Paul Popelier by email ([Email Address Removed]). Please note that to apply for this studentship you must submit the relevant information via the University’s online application form.