Seminar 30th November 2011 1 p.m. 27/2003
Overcoming frustration with local elevation: enhancing the sampling in biomolecular simulations biased with experimental data
Dr Jane Allison
- Bayesian probability, Bioinformatics, Biomathematics, Biomechanics, Biomedical, Biomolecular simulations, CASTEP, Complex Systems, Computer Science, Developmental Biology, Economics, Evolutionary Algorithms, FFT, Finite differences, Finite elements, Formal methods, GPU, HECToR, HPCx, Iridis, Lyceum, Mathematica, Matlab, Molecular Dynamics, Molecular Mechanics, Monte Carlo, MPI, Multi-physics, Multi-scale, Multigrid solvers, Multipole methods, NWCHEM, Onetep, Optimisation, Quantum Chemistry, SVN, VTK, Windows
- Chris-Kriton Skylaris
To determine or refine structures of biological macromolecules, experimental data is converted into structural properties such as inter-atomic distances or dihedral angles, and incorporated into a modelling program. The relationships used to convert between experimental data and structural parameters can introduce significant uncertainty, however. This is exacerbated by the fact that experimental measurements are typically averages over time and over a large number of molecules. If the molecule is flexible, a wide distribution of different values contribute to the measured average. The shape of this distribution cannot be specified solely from knowledge of the average. Molecular dynamics (MD) provides a means of generating ensembles of structures that is free from a priori assumptions about the nature of the underlying distribution. However, MD can suffer from limited sampling with respect to time and conformational space. Moreover, the quality of the structures is dependent on the choice of force-field and solvation model. I will explain and illustrate these problems using NMR data describing side-chain dihedral angles. In addition, I will introduce a method for biomolecular structure refinement based on adaptive restraints that provides an elegant means of overcoming many of these problems.