Bates lab

Biomolecular Modelling Laboratory

 Areas of interest

Computer simulation of crowded cell interior

Within Biomolecular Modelling we study fundamental and challenging problems in both structural and systems biology; in particular, how macromolecules interact at the atomic level to facilitate cellular events. Much of the work involves the design of novel computer algorithms that are based upon the principles of physics and evolutionary biology. These simulations are proving to be important in helping to interpret experimental data and suggest further experiments to probe complex molecular systems. Outlined below are two systems currently under investigation.

Characterising changes in the rate of protein-protein dissociation upon interface mutation

Predicting the effects of mutations on the kinetic rate constants of protein-protein interactions is important to both the modelling of complex diseases, such as cancer, and the design of effective protein inhibitors. We have designed a novel computational approach to relate the changes in protein complex dissociation rates (koff) upon mutation to the energetics and architecture of hotspots; hotspots refer to a subset of residues at the protein-protein interface, which are able to significantly destabilise the binding free energy by more than 2 kcal/mol when mutated to alanine. We have constructed a number of machine learning models to take account of both molecular and hotspot descriptors when predicting changes in koff that correlates well with experimentally determined off rates, and moreover, can identify rare stabilising mutations, important for the rational design of high protein-protein binding affinities.

Schematic diagram of the approach taken to estimate changes in koff for mutations at a protein-protein interface; example mutation, Arginine (Arg) to Leucine (Leu). To facilitate the calculation of the correct energetics, a computational alanine scan is performed at the interface pre- and post-mutation. For more details see Agius et al., 2013; PLoS Computer Biol. 9(9): e1003216.

Figure 1: Schematic diagram of the approach taken to estimate changes in koff for mutations at a protein-protein interface; example mutation, Arginine (Arg) to Leucine (Leu). To facilitate the calculation of the correct energetics, a computational alanine scan is performed at the interface pre- and post-mutation. For more details see Agius et al., 2013; PLoS Computer Biol. 9(9): e1003216.

Multiscale modelling of cancer cell motility

Cell motility is required for many biological processes, including cancer metastasis. However, predicting the optimal migration strategy or the effects of experimental perturbation for a migrating cancer cell is difficult. Hence we have constructed a computational model for cancer cell motility. In collaboration with Eric Sahai (Tumour Cell Biology Group), experimental data on cancer cell morphology and dynamics was utilised in the construction of the model, and the predictions of our model validated against in vivo data. The model is being used to probe the more effective combinations of biochemical interventions aimed at reducing cancer cell motility (Tozluoglu et al., 2013; Nat Cell Biol. 15(7): 751-62).

Validating biomarkers for clear cell renal carcinoma

Candidate biomarkers have been identified for clear cell renal carcinoma (ccRCC) patients, but most have not been validated. In collaboration with Charles Swanton (Translational Cancer Therapeutics Laboratory) we have analysed 28 genetic or transcriptional biomarkers in 350 ccRCC patients in terms of cancer-specific survival (CSS). Our conclusion from the study is that only one biomarker, a gene expression set called ccB, could be considered to be an independent prognostic biomarker for CSS (see Gulati et al. for more details).

Mapping the shape of protein-protein binding funnels with SwarmDock

Predicting the effects of mutations on the kinetic rate constants of protein-protein interactions is important to both the modelling of complex diseases, such as cancer, and the design of effective protein inhibitors. To facilitate our understanding of how mutations affect binding kinetics we are mapping the conformational shapes of binding funnels for wild type and mutated binding partners. We are developing software to display the output of our publicly available macromolecular docking program SwarmDock to interpret these binding funnel shapes (see Figure 2). Details of how to effectively use our docking program are given in a recent publication (Torchala and Bates 2014).

Schematic diagram of the binding funnel between two proteins, actin and one of its binding partners, a vitamin D-binding protein. The left hand panel shows the complete search space between the two proteins as a connected graph of conformational states. Larger nodes represent more stable protein conformations. The highly connected set of nodes represents the true binding funnel. The right hand panel shows an exploded view of the docking funnel, actin in magenta, and in green the final conformation of the vi

Figure 2: Schematic diagram of the binding funnel between two proteins, actin and one of its binding partners, a vitamin D-binding protein. The left hand panel shows the complete search space between the two proteins as a connected graph of conformational states. Larger nodes represent more stable protein conformations. The highly connected set of nodes represents the true binding funnel. The right hand panel shows an exploded view of the docking funnel, actin in magenta, and in green the final conformation of the vitamin D-binding protein. Moving from A (edge of the binding funnel) to D (near the final bound state) stabilising interactions between the two proteins can be seen to increase.