| Summary: | Proteins are the essential units of biological processes, but modelling their dynamics is a very computationally expensive task. A wide variety of simulation techniques exist, a popular example being Molecular Dynamics. However, such models typically involve detailed simulation of the protein's structure at or near the atomic level and as such are unsuitable for modelling biological systems composed of large or multiply interacting proteins. This research takes a coarse-graining approach, called Fluctuating Finite Element Analysis, in which large, globular proteins are approximated by viscoelastic continua subject to thermal noise. Each protein is discretised into a tetrahedral mesh, parameterised locally by its bulk continuum properties. The forces are then calculated using Finite Element Analysis. A parallel implementation of the FFEA algorithm has been developed for use on high performance computing facilities. The scalability of the algorithm with respect to number of cores and system size, and its stability with respect to integration time step has been investigated. A pipeline for fully automated FFEA system creation from atomistic (X-ray crystallography and NMR) or low resolution data (cryo-EM and SAXS) has also been developed. In order to tackle multiprotein systems, the FFEA model has been extended to include van derWaals interactions and electrostatics. FFEA has been applied to a number of diverse biological systems. The van der Waals scheme was tested through simulation of myoglobins interacting with a polystyrene substrate. The major modes of motion of V- and A- type rotary ATPases were extracted using Principal Component Analysis, and compared with the normal modes obtained from the Elastic Network Model. Finally, the effect of axonemal dynein c's interaction with the microtubule track on its step length and exploration of binding sites was investigated. A mapping was developed to allow in-simulation conformational switching of the dynein motor.
|
|---|
