Unravelling the molecular mechanisms that underpin regulatory kinase function using new computational tools

The aim of this project is to develop a new computational tool for modelling large scale conformational change in biomolecules. The tool will be used to characterise the molecular mechanism of kinase (in)activation. In collaboration with Johnson & Johnson, the in silico model will be tested at bench level for further insight and model refinement. The proposed tool will build on recent software developments in the computational biophysics group of Dr Bryce. At the interface of computer science and biophysics, we have implemented a method[1] which combines the power of swarm intelligence methods with molecular dynamics simulation to significantly accelerate conformational searching. Applying this method for the first time to biomolecular systems, our tests on peptide and mini-protein folding suggest five-fold increases in efficiency over a comparable method (manuscript in preparation). The success of the approach lies in mimicking the cooperative behaviour of birds flocking in how multiple copies of the system interact[2], leading to smoothing of the energy landscape and avoidance of local minima traps. We see the opportunity to adapt this tool to enhance the phase space volume sampled, by suitably modifying the coupling between system replicas. From these software modifications (Years 1-2 of the project), we will more accurately and cheaply model the structural and energetic features of large scale conformational changes, where the energy landscape is rough and barriers can be large compared to kT. This tool will then be applied to study of kinase conformation and function. Kinases are part of many vital signalling cascades and are exquisitely regulated. These proteins can adopt several conformations, of which two extremes are most commonly found: an open, active state and a closed, inactive state. These states feature a conserved Asp-Phe-Gly (DFG) motif within the activation loop: the active catalytic state has a DFG-in conformation, whereas a DFG-out conformation is not optimal for binding ATP substrate. To date, the pathway, energetics and intermediates between these states are not well understood. Neither is it known why some kinases do not appear to adopt a DFG-out conformation. This project will therefore apply this computational methodology to accurately model the energetics and structure of this large scale conformational rearrangement for selected kinases. DFG-in to DFG-out transitions will be driven using umbrella and swarm potentials for four kinases where both states are known (VEGFR2, c-Src, p38, Abl) and for three kinases which are known not to adopt a DFG-out structure (GSK3b, CDK1, one in-house kinase target). Hypothetical inactive structures for the latter kinases will be modelled by analogy (by the student at Johnson & Johnson during the first 3 months of the project). From these kinase simulations, we will identify key amino acids and molecular motifs along the transition pathway that could act as potential points of molecular recognition. We will then apply computer-aided design methods to screen commercial and Johnson & Johnson libraries for potential small molecules that can specifically recognise these protein conformations (Years 3-4). To provide validation for these in silico predictions, as well as a basis for further refinement of the model itself, promising small molecule hits will be assayed at Johnson & Johnson for kinase inhibitory activity, with the potential for further characterisation by structural biology and biophysical methods. Therefore, this interdisciplinary project seeks to develop a predictive computational tool to answer the fundamental questions of what energetic, structural and dynamic features underpin the active-to-inactive transition of many kinases and discern the underlying reason why some kinases do not adopt a DFG-out conformation. [1] Huber et al, J. Phys. Chem. A 1998, 102, 5943. [2] Kennedy et al, Proc. IEEE Intl. Conf. Neural Networks 1995, 4, 1942.